KR20220165111A - Method for classification of medical image - Google Patents

Abstract

A method for classifying medical images performed by a computing device according to one embodiment of the present disclosure is disclosed. The method includes the steps of: inputting at least one medical image to a neural network model based on unsupervised learning and estimating the input suitability for a model which reads the medical image; and selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability.

Description

Method for classifying medical images {METHOD FOR CLASSIFICATION OF MEDICAL IMAGE}

The present invention relates to a method for processing a medical image, and more particularly, to a method for classifying a medical image suitable for an input of a reading model.

Medical images are data that enable us to understand the physical state of various organs of the human body. Medical images include digital radiographic imaging (X-ray), computed tomography (CT), or magnetic resonance imaging (MRI).

In accordance with the need to easily understand a person's condition by analyzing medical images, various models for reading medical images based on deep learning are being developed. In general, a model that reads a medical image according to a design tailored to a specific domain analyzes the medical image and provides information necessary for determining a human condition based on the analysis result. Therefore, in order to accurately read the reading model, it is necessary to input a medical image suitable for the distribution of data used for learning the reading model to the reading model.

For example, in the case of a model that analyzes bone age based on human skull bones, a user-desired output value can be accurately obtained only when a medical image of a hand that fits the learning distribution of the model is input to the model. If a medical image of a body part other than a hand, such as a foot, is input, a model that analyzes bone age based on the human skull cannot help but output an inaccurate analysis result. In this way, data that deviate from the learning distribution of the model and affect the performance of the model is generally referred to as OOD (out-of-distribution) in the field of deep learning.

US Patent Publication No. 2020-0211187 (July 2, 2020) discloses a system and method for detecting centers of ossification and assessing bone age.

The present disclosure has been made in response to the above background art, and by detecting data corresponding to OOD (out-of-distribution) of the reading model to ensure stability and performance of the reading model, medical images suitable as inputs of the reading model Its purpose is to provide a method for selecting

A method for classifying medical images performed by a computing device according to an embodiment of the present disclosure for realizing the above object is disclosed. The method may include inputting at least one medical image to a neural network model based on unsupervised learning and estimating the suitability of the input to a model for reading the medical image; and selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability.

In an alternative embodiment, estimating the suitability of an input to a model for reading the medical image may include extracting a feature of the at least one medical image using the neural network model; mapping the extracted feature to a hypersphere of a latent space using the neural network model; and estimating input suitability for a model for reading the medical image of the at least one medical image based on a feature mapped to the hypersphere of the latent space using the neural network model.

In an alternative embodiment, the step of estimating input suitability for a model for reading the medical image of the at least one medical image based on a feature mapped to a hypersphere of the latent space includes: using the neural network model; The method may include estimating input suitability of the at least one medical image for a model that reads the medical image, based on a distance between a center of the supersphere of the latent space and a feature mapped to the supersphere of the latent space.

In an alternative embodiment, the center of the supersphere of the latent space may correspond to an average of features expressed in the latent space of all medical images input to the neural network model.

In an alternative embodiment, the step of selecting the input data of the model for reading the medical image may include comparing the estimated input suitability with a predetermined threshold, and from the at least one medical image, the model for reading the medical image. It may include the step of selecting the input data of.

In an alternative embodiment, the input data of the model that reads the medical image is an image that does not include a body part targeted by the model that reads the medical image, and a reading criterion for the model that reads the medical image. It may be a medical image that is not at least one of an image that is not captured at the position or angle at which the medical image is captured, or an image that includes a quality factor that interferes with the interpretation of a model that interprets the medical image.

In an alternative embodiment, the neural network model may perform data augmentation on a medical image input for learning, and may be pre-learned through unsupervised learning based on the medical image on which the data augmentation is performed. .

In an alternative embodiment, the neural network model extracts features of a medical image input for learning, and maps features of a medical image generated through data augmentation among the extracted features to be far from the center of the supersphere in the latent space. It can be pre-learned through unsupervised learning. In this case, the features of the medical image generated through the data augmentation may be mapped so as to be farther from the center of the supersphere in the latent space as the magnitude of the data augmentation increases.

In an alternative embodiment, the data augmentation may include data augmentation based on geometry transformation, data augmentation by applying contrast to medical images, data augmentation by applying noise to medical images, or medical data augmentation by applying noise to medical images. It may include at least one of data augmentation applying an artifact or a replacement image to at least a portion of the image.

In an alternative embodiment, the model for reading the medical image may include a model for evaluating bone age of a body part included in the medical image.

A computer program stored in a computer readable storage medium is disclosed according to an embodiment of the present disclosure for realizing the above object. When the computer program is executed on one or more processors, the following operations for classifying medical images are performed, and the operations include: inputting at least one medical image to a neural network model based on unsupervised learning, so as to classify the medical image. estimating input suitability for the model being read; and selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability.

According to an embodiment of the present disclosure for realizing the above object, a computing device for classifying medical images based on medical images is disclosed. The device may include a processor including at least one core; a memory containing program codes executable by the processor; and a network unit for receiving at least one medical image, wherein the processor inputs the at least one medical image to a neural network model based on unsupervised learning, estimates input suitability for a model that interprets the medical image, and , Input data of a model for reading the medical image may be selected from the at least one medical image based on the estimated input suitability.

The present disclosure may provide a method of selecting a suitable medical image as an input of a reading model by sensing data corresponding to out-of-distribution (OOD) of the reading model to ensure stability and performance of the reading model.

1 is a block diagram of a computing device for classifying medical images according to an embodiment of the present disclosure.
2 is a schematic diagram showing a neural network according to an embodiment of the present disclosure.
3 is a block diagram illustrating a process of processing a medical image of a neural network model according to an embodiment of the present disclosure.
4 is a block diagram showing the structure and calculation process of a neural network model according to an embodiment of the present disclosure.
5 is a conceptual diagram illustrating data augmentation of a bone image and criteria for an image unsuitable for an input of a reading model according to an embodiment of the present disclosure.
6 is a flowchart illustrating a learning process of a neural network model according to an embodiment of the present disclosure.
7 and 8 are flowcharts illustrating a process of classifying medical images of a computing device according to an embodiment of the present disclosure.
9 is a schematic diagram of a computing environment according to one embodiment of the present disclosure.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific details.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an execution of software. For example, a component may be, but is not limited to, a procedure, processor, object, thread of execution, program, and/or computer running on a processor. For example, both an application running on a computing device and a computing device may be components. One or more components may reside within a processor and/or thread of execution. A component can be localized within a single computer. A component may be distributed between two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. Components may be connected, for example, via signals with one or more packets of data (e.g., data and/or signals from one component interacting with another component in a local system, distributed system) to other systems and over a network such as the Internet. data being transmitted) may communicate via local and/or remote processes.

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless otherwise specified or clear from the context, “X employs A or B” is intended to mean one of the natural inclusive substitutions. That is, X uses A; X uses B; Or, if X uses both A and B, "X uses either A or B" may apply to either of these cases. Also, the term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements, and/or groups thereof. Also, unless otherwise specified or where the context clearly indicates that a singular form is indicated, the singular in this specification and claims should generally be construed to mean "one or more".

In addition, the term “at least one of A or B” should be interpreted as meaning “when only A is included”, “when only B is included”, and “when A and B are combined”.

Those skilled in the art will further understand that the various illustrative logical blocks, components, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented using electronic hardware, computer software, or combinations of both. It should be recognized that it can be implemented as To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented in hardware or as software depends on the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The description of the presented embodiments is provided to enable any person skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure. The general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present invention is not limited to the embodiments presented herein. The present invention is to be accorded the widest scope consistent with the principles and novel features set forth herein.

In the present disclosure, network functions, artificial neural networks, and neural networks may be used interchangeably.

On the other hand, the term "image" used throughout the detailed description and claims of the present invention refers to multidimensional data composed of discrete image elements (eg, pixels in a two-dimensional image). In other words, the term “image” refers to either a visible object (e.g., displayed on a video screen) or a digital representation of that object (e.g., a file corresponding to the pixel output of a CT, MRI detector, etc.) could be a term.

For example, "image" may refer to blood collected by computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, or any other medical imaging system known in the art. It may be an image of a subject. The image does not necessarily have to be provided in a medical context, but may be provided in a non-medical context, such as X-ray imaging for security screening.

Throughout the detailed description and claims of the present invention, the 'DICOM (Digital Imaging and Communications in Medicine)' standard is a term that collectively refers to various standards used for digital image expression and communication in medical devices, The DICOM standard is published by a joint committee formed by the American College of Radiology (ACR) and the National Electrical Manufacturers Association (NEMA).

In addition, throughout the detailed description and claims of the present invention, 'Picture Archiving and Communication System (PACS)' is a term that refers to a system that stores, processes, and transmits according to the DICOM standard, and includes X-ray, CT, , Medical imaging images acquired using digital medical imaging equipment such as MRI are stored in DICOM format and can be transmitted to terminals inside and outside the hospital through a network, and reading results and medical records can be added to them.

1 is a block diagram of a computing device for classifying medical images according to an embodiment of the present disclosure.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In one embodiment of the present disclosure, the computing device 100 may include other components for performing a computing environment of the computing device 100, and only some of the disclosed components may constitute the computing device 100.

The computing device 100 may include a processor 110 , a memory 130 , and a network unit 150 .

The processor 110 may include one or more cores, and includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. unit), data analysis, and processors for deep learning. The processor 110 may read a computer program stored in the memory 130 and process data for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning a neural network. The processor 110 is used for neural network learning, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating neural network weights using backpropagation. calculations can be performed. At least one of the CPU, GPGPU, and TPU of the processor 110 may process learning of the network function. For example, the CPU and GPGPU can process learning of network functions and data classification using network functions. In addition, in an embodiment of the present disclosure, the learning of a network function and data classification using a network function may be processed by using processors of a plurality of computing devices together. In addition, a computer program executed in a computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

According to an embodiment of the present disclosure, the processor 110 may select a medical image suitable as input data of a model that reads the medical image. A model that reads medical images needs to receive medical images included in a learning distribution in order to derive accurate calculation and analysis results based on the medical images. That is, when a medical image deviating from the learning distribution of the reading model is input to the reading model, the model reading the medical image cannot derive an appropriate output value desired by the user. Accordingly, it is necessary to determine whether the medical image is suitable for input to the model for reading the medical image before inputting the medical image to the model for reading the medical image. According to such a need, the processor 110 aims to preliminarily filter out medical images unsuitable as a reading model from among medical images and select medical images suitable as a reading model in advance.

The processor 110 may estimate the suitability of the input to the interpretation model for the medical image through the neural network model. Here, the input suitability may be understood as an index indicating how much a medical image is included in (or deviated from) the learning distribution of the reading model. Input suitability may be understood as an analysis value output as an operation result of a neural network model indicating how suitable a medical image is as input data of a reading model. For example, the processor 110 may input at least one medical image acquired through the network unit 150 to a neural network model and estimate input suitability for a model that reads a bone image. In this case, the medical image may be a bone image or an image of a body part other than the bone (eg, skull, etc.). In addition, although the medical image is a bone image, it may be an image generated with low quality because the target region is not accurately photographed or there are interfering factors. With respect to the medical images generated according to various types or conditions, the processor 110 may estimate the suitability of an input to a model that reads a bone image through a neural network model. The input suitability of the medical image may be a score value representing the degree to which the medical image is suitable for the input of a model that reads a bone image.

The processor 110 may select suitable data as input data of the reading model from a group of medical images whose input suitability is estimated based on the input suitability of the medical image estimated through the neural network model. When the input suitability of the plurality of medical images is estimated, the processor 110 determines whether the images can be individually used as input data of the interpretation model based on the input suitability of each of the plurality of medical images, and Input data of the reading model may be selected from images. When input suitability for a single image is estimated, the processor 110 may determine whether the corresponding image can be used as input data of a reading model based on the input suitability of the corresponding image, and determine whether to use the corresponding image. For example, the processor 110 may determine whether the medical image is suitable as input data for a model that reads a bone image by comparing input suitability of a medical image estimated through a neural network model with a predetermined threshold value. When one of the plurality of medical images corresponds to a bone image and the input suitability is equal to or greater than a predetermined threshold, the processor 110 determines that the image is suitable as input data for a model that reads the bone image, and selects the corresponding image as input data. can If one of the plurality of medical images corresponds to an image of a body part other than the carpal bones (e.g. skull, etc.) and the input suitability is less than a predetermined threshold, the processor 110 converts the corresponding image to a model that reads the carpal bones image. It can be judged that the data is not suitable. In this case, an image determined to be unsuitable as input data of a model that reads a bone image may be classified as an out-of-distribution (OOD) of a model that reads a bone image.

Meanwhile, the above-described threshold-based determination is only an example. Accordingly, selection of input data by the processor 110 may be performed based on threshold-based judgment or may be performed through a separate neural network model. That is, selection of input data by the processor 110 may be variously performed through any one or combination of a rule-based operation or a learning-based operation within a range understandable by those skilled in the art.

According to an embodiment of the present disclosure, the memory 130 may store any type of information generated or determined by the processor 110 and any type of information received by the network unit 150 .

According to an embodiment of the present disclosure, the memory 130 is a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory (eg, SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) -Only Memory), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium. The computing device 100 may operate in relation to a web storage that performs a storage function of the memory 130 on the Internet. The above description of the memory is only an example, and the present disclosure is not limited thereto.

The network unit 150 according to an embodiment of the present disclosure may use any type of known wired or wireless communication system.

The network unit 150 may receive a medical image representing body organs from a medical image storage and transmission system or a medical image capture system. For example, a medical image in which body organs are expressed is a medical image obtained by photographing an upper limb, and may be training data or inference data of a neural network model. In this case, the medical image in which the body organs are expressed may include all images related to the body organs obtained through imaging, such as X-ray images and CT images.

In addition, the network unit 150 may transmit and receive information processed by the processor 110, a user interface, and the like through communication with other terminals. For example, the network unit 150 may provide a user interface generated by the processor 110 to a client (eg, a user terminal). In addition, the network unit 150 may receive an external input of a user authorized as a client and transmit it to the processor 110 . At this time, the processor 110 may process operations such as outputting, correcting, changing, adding, and the like of information provided through the user interface based on the user's external input received from the network unit 150 .

Meanwhile, the computing device 100 according to an embodiment of the present disclosure may include a server as a computing system that transmits and receives information through communication with a client. In this case, the client may be any type of terminal capable of accessing the server. For example, the computing device 100 as a server may receive a medical image from a medical imaging terminal, predict a disease, and provide a user interface including the predicted result to the user terminal. At this time, the user terminal may output a user interface received from the computing device 100 as a server, and receive or process information through interaction with the user.

In an additional embodiment, the computing device 100 may include any type of terminal that receives data resources generated by any server and performs additional information processing.

2 is a schematic diagram showing a neural network according to an embodiment of the present disclosure.

A neural network model according to an embodiment of the present disclosure may include a neural network for estimating the suitability of an input of a medical image to a model for reading a medical image. Throughout this specification, neural network, network function, and neural network may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes. Nodes (or neurons) constituting neural networks may be interconnected by one or more links.

In a neural network, one or more nodes connected through a link may form a relative relationship of an input node and an output node. The concept of an input node and an output node is relative, and any node in an output node relationship with one node may have an input node relationship with another node, and vice versa. As described above, an input node to output node relationship may be created around a link. More than one output node can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of data of the output node may be determined based on data input to the input node. Here, a link interconnecting an input node and an output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value may be determined based on the weight.

As described above, in the neural network, one or more nodes are interconnected through one or more links to form an input node and output node relationship in the neural network. Characteristics of the neural network may be determined according to the number of nodes and links in the neural network, an association between the nodes and links, and a weight value assigned to each link. For example, when there are two neural networks having the same number of nodes and links and different weight values of the links, the two neural networks may be recognized as different from each other.

A neural network may be composed of a set of one or more nodes. A subset of nodes constituting a neural network may constitute a layer. Some of the nodes constituting the neural network may form one layer based on distances from the first input node. For example, a set of nodes having a distance of n from the first input node may constitute n layers. The distance from the first input node may be defined by the minimum number of links that must be passed through to reach the corresponding node from the first input node. However, the definition of such a layer is arbitrary for explanation, and the order of a layer in a neural network may be defined in a method different from the above. For example, a layer of nodes may be defined by a distance from a final output node.

An initial input node may refer to one or more nodes to which data is directly input without going through a link in relation to other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. Also, the hidden node may refer to nodes constituting the neural network other than the first input node and the last output node.

In the neural network according to an embodiment of the present disclosure, the number of nodes in the input layer may be the same as the number of nodes in the output layer, and the number of nodes decreases and then increases again as the number of nodes progresses from the input layer to the hidden layer. can In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes of the input layer may be less than the number of nodes of the output layer and the number of nodes decreases as the number of nodes increases from the input layer to the hidden layer. have. In addition, the neural network according to another embodiment of the present disclosure is a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the number of nodes increases from the input layer to the hidden layer. can A neural network according to another embodiment of the present disclosure may be a neural network in the form of a combination of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks can reveal latent structures in data. In other words, it can identify the latent structure of a photo, text, video, sound, or music (e.g., what objects are in the photo, what the content and emotion of the text are, what the content and emotion of the audio are, etc.). . Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted boltzmann machines (RBMs). machine), a deep belief network (DBN), a Q network, a U network, a Siamese network, a Generative Adversarial Network (GAN), and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In one embodiment of the present disclosure, the network function may include an autoencoder. An autoencoder may be a type of artificial neural network for outputting output data similar to input data. An auto-encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between input and output layers. The number of nodes of each layer may be reduced from the number of nodes of the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically with the reduction from the bottleneck layer to the output layer (symmetrical to the input layer). Autoencoders can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to dimensions after preprocessing of input data. In the auto-encoder structure, the number of hidden layer nodes included in the encoder may decrease as the distance from the input layer increases. If the number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) is too small, a sufficient amount of information may not be conveyed, so more than a certain number (e.g., more than half of the input layer, etc.) ) may be maintained.

The neural network may be trained using at least one of supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Learning of the neural network may be a process of applying knowledge for the neural network to perform a specific operation to the neural network.

A neural network can be trained in a way that minimizes output errors. In the learning of the neural network, the learning data is repeatedly input into the neural network, the output of the neural network for the training data and the error of the target are calculated, and the error of the neural network is transferred from the output layer of the neural network to the input layer in the direction of reducing the error. It is a process of updating the weight of each node of the neural network by backpropagating in the same direction. In the case of supervised learning, each learning data is labeled with the correct answer (ie, labeled learning data), and in the case of unsupervised learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of supervised learning related to data classification may be data in which each learning data is labeled with a category. Labeled training data is input to a neural network, and an error may be calculated by comparing an output (category) of the neural network and a label of the training data. As another example, in the case of unsupervised learning for data classification, an error may be calculated by comparing input learning data with a neural network output. The calculated error is back-propagated in a reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back-propagation. The amount of change in the connection weight of each updated node may be determined according to a learning rate. The neural network's computation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of iterations of the learning cycle of the neural network. For example, a high learning rate may be used in the early stage of neural network training to increase efficiency by allowing the neural network to quickly obtain a certain level of performance, and a low learning rate may be used in the late stage to increase accuracy.

In neural network learning, generally, training data can be a subset of real data (ie, data to be processed using the trained neural network), and therefore, errors for training data are reduced, but errors for real data are reduced. There may be incremental learning cycles. Overfitting is a phenomenon in which errors for actual data increase due to excessive learning on training data. For example, a phenomenon in which a neural network that has learned a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing the error of machine learning algorithms. Various optimization methods can be used to prevent such overfitting. To prevent overfitting, methods such as increasing the training data, regularization, inactivating some nodes in the network during learning, and using a batch normalization layer should be applied. can

3 is a block diagram illustrating a process of processing a medical image of a neural network model according to an embodiment of the present disclosure.

Referring to FIG. 3 , the neural network model 200 according to an embodiment of the present disclosure is based on at least one medical image 11, in which a medical image 11 is included in the learning distribution of a model that reads a specific image. An input suitability 12 representing the degree can be generated. The neural network model 200 may receive at least one medical image 11 and generate an input fitness 12 that is an index indicating whether the medical image 11 is suitable for an input of a model that reads a specific image. For example, the neural network model 200 may receive the medical image 11 and express features of the medical image 11 in a latent space. The neural network model 200 is based on the correlation between the features of the medical image 11 expressed in the latent space and the features of other images (or pre-learned features) expressed in the latent space. The input fitness 12 of the medical image 11 expressed in can be calculated. In other words, the neural network model 200 may calculate the input fitness 12 of the medical image 11 expressed in the latent space based on the distribution of features of the images expressed in the latent space. In this case, the input fitness 12 may be a score value representing a probability that the medical image 11 corresponds to an out-of-distribution (OOD) of a model that reads a specific medical image. Input relevance 12 may be used interchangeably with the expression input relevance. In this way, in order for the neural network model 200 to be able to calculate the input fitness 12 of the medical image 11, the neural network model 200 uses a criterion on the latent space as input data of the model for reading a specific medical image. Based on the features, it can be learned in advance through unsupervised learning.

A reading model serving as a standard for the input fitness 12 output by the neural network model 200 may be a machine learning-based model that reads a specific medical image. The reading model may be a pre-learned model to receive and read a specific image, and output a reading result of the specific image. Therefore, in order to guarantee the performance of the reading model, images included in the training distribution of the reading model should be input to the reading model. For example, when an input image of the interpretation model is an image that does not include a body part targeted by the interpretation model, it is difficult for the interpretation model to output an accurate reading result. If the input image of the reading model is an image that is not captured at a position or angle that is a reference for reading the reading model, it is difficult for the reading model to output an accurate reading result as well. In addition, even when an input image of the interpretation model is an image including a quality factor that hinders the interpretation of the model for reading the medical image, the interpretation model is similarly difficult to output an accurate reading result. That is, images like the above example correspond to out-of-distribution (OOD) images out of the learning distribution of the reading model. That is, in order to prevent such an out-of-distribution (OOD) image of the interpretation model from being input to the interpretation model in advance, the neural network model 200 calculates the input suitability 12 of the medical image 11 to the interpretation model. It is possible to perform an operation that calculates.

4 is a block diagram showing the structure and calculation process of a neural network model according to an embodiment of the present disclosure.

Referring to FIG. 4 , the neural network model 200 according to an embodiment of the present disclosure may extract features 22 of a medical image 21 through an encoder 210 . The neural network model 200 may sequentially input the medical images 21 to the encoder 210 to extract features 22 of each of the medical images 21 . The neural network model 200 may convert the feature 22 of the medical image 21 extracted through the encoder 210 from the representation of the existing feature space (A) to the representation of the latent space (B). In other words, the neural network model 200 may map the feature 22 of the medical image 21 extracted through the encoder 210 onto the latent space B. At this time, the neural network model 200 may map the feature 22 extracted through the encoder 210 to the hypersphere of the latent space B. The neural network model 200 may estimate input suitability for the interpretation model of the medical image 11 based on the feature 22 mapped to the hypersphere of the latent space B. The first phrase of the latent space B can be understood as a kind of criterion for determining the suitability of a learned input based on the distribution of features in the latent space B in the unsupervised learning process of the neural network model 200 .

For example, the neural network model 200 is based on the distance between the center (c) of the hypersphere of the latent space (B) and the feature 22 mapped to the hypersphere of the latent space (B) for the medical image 21. The fit of the input into the reading model can be estimated. In this case, the center c of the hypersphere in the latent space B may correspond to the average of features expressed in the latent space B of all medical images input to the neural network model 200 . The neural network model 200 calculates the value of the medical image 21 based on how far apart the features of the medical image 21 to determine input suitability are based on the average of the features expressed in the latent space B through prior learning. We can estimate the input suitability for . For this estimation, the neural network model 200 includes most of the features of the images suitable for the input of the reading model among the medical images input for learning, but minimizes the radius of the hypersphere (R) of the latent space (B) It can be pre-learned based on unsupervised learning. In other words, for the aforementioned estimation, the neural network model 200 selects the features of images that are not suitable for the input of the interpretation model among the medical images input for learning at the center of the hypersphere (c) of the latent space (B) (or the latent space B). It can be pre-learned through unsupervised learning that maps away from the average of the features represented in space (B).

When the model for reading a medical image according to an embodiment of the present disclosure is a model for evaluating a bone age by reading a frontal medulla bone image, the neural network model 200 maps the medical image 21 to the supernatant of the latent space B. ), a score representing the suitability of an input to a model for evaluating bone age can be calculated by reading the frontal medulla bone image. The score indicating the suitability of the input to the reading model may correspond to the distance between the center c of the hypersphere of the latent space B and the feature of the medical image 21 mapped to the hypersphere of the latent space B. At this time, the neural network model 200 reads the frontal medulla image and moves the features of the medical images corresponding to the out-of-distribution (OOD) of the model for evaluating the bone age away from the center of the supersphere in the latent space B. It can also be pre-learned based on learning. In addition, the neural network model 200 prevents features of medical images corresponding to out-of-distribution (OOD) of a model for evaluating bone age by reading a frontal medulla image to be maximally included in the supersphere of the latent space (B). It can be pre-learned based on unsupervised learning. Medical images corresponding to out-of-distribution (OOD) of a model for evaluating bone age by reading a frontal medulla image may include the following images in [Table 1]. However, since [Table 1] is only an example, the out-of-distribution (OOD) of the model that evaluates bone age by reading the frontal medulla bone image is configured in various ways according to the learning distribution of the model in addition to [Table 1]. It can be.

1. Images that do not include the carpal bones that the model targets
(1) Images taken of parts other than the hand
(2) Images with the other hand on top of the patient's hand
(3) Video taken with both hands, not one
2. Images that are not filmed at a location or angle that is the standard for model reading
(1) Images taken in a view position other than the frontal carpal bone
(2) Images taken with excessive rotation instead of frontal
3. Images containing quality factors that interfere with model interpretation
(1) Images covering or affecting bone and joint parts for reading
(2) Image containing severe noise
(3) Image with wrong contrast

5 is a conceptual diagram illustrating data augmentation for a bone image according to an embodiment of the present disclosure and criteria for an image unsuitable for an input of an interpretation model. The neural network model 200 according to an embodiment of the present disclosure is busy It is very important to reduce the bias of data for learning because it is pre-learned based on pre-learning. For example, when medical images for learning input to the neural network model 200 are composed of only images included in the learning distribution of the interpretation model (or suitable for the input of the interpretation model), the neural network model 200 is It can be overfitted based on the features that fit the input. In this case, a problem may occur in that the neural network model 200 cannot properly perform an operation on an input image for determining input suitability. Due to the nature of the domain of medical images, it is not realistically easy to obtain various types of learning data that are not biased to one side, so the problems described above can easily occur in the unsupervised learning process.

In order to solve the above problem, the neural network model 200 may perform data augmentation on the medical image input for learning. That is, in order to solve the overfitting problem that can easily occur in the process of learning the input suitability for the interpretation model of medical images due to unsupervised learning and domain characteristics, the neural network model 200 is suitable for the interpretation model (or the interpretation model Based on the images included in the learning distribution of , data augmentation may be performed to create images in which the reading model is inappropriate (or out of the learning distribution of the reading model). In this case, the data augmentation may include data augmentation based on geometry transformation, data augmentation by applying contrast to the medical image, data augmentation by applying noise to the medical image, or artifacts or replacement images in at least a part of the medical image. It may include at least one of data augmentation applying . Through the data augmentation of the example as described above, the neural network model 200 generates a medical image corresponding to the OOD (out-of-distribution) of the reading model on its own and uses it for unsupervised learning, thereby overfitting and learning data The bias problem of can be easily solved.

Referring to FIG. 5 , when the model for reading a medical image according to an embodiment of the present disclosure is a model that evaluates bone age by reading a frontal medulla bone image, the neural network model 200 for unsupervised learning is a model for reading Data augmentation may be performed based on the target frontal carpal bone image 30 . The neural network model 200 obtains an image 41 representing a body part other than the carpal bone from the frontal carpal tunnel image 30 through data augmentation, an image 42 in which a part of the carpal bone is obscured by an obstruction, and a carpal bone of another patient. OOD (out -of-distribution) You can create an image. Specifically, the neural network model 200 generates an image 41 representing body parts other than the carpal bone from the frontal carpal tunnel image 30 through geometric transformation-based data augmentation (e.g. rotate, shearX, shearY, translateX, translateY, etc.). and at least one of an image 46 rotated at an angle other than the front. The neural network model 200 generates an image 44 containing noise through data augmentation (e.g. autocontrast, equalize, brightness, gaussian noise, salt and pepper noise, laplacian noise, etc.) applying contrast or noise to medical images. can In addition, the neural network model 200 uses data augmentation (e.g. cut-out, cut-mix, mosaic data augmentation, etc.) to apply artifacts or substitute images to at least a part of a medical image, so that a part of the carpal tunnel is obscured by an obstacle (42). ), an image 43 including the carpal bones of another patient, or an image 45 including two carpal bones may be generated.

The neural network model 200 may perform unsupervised learning in which features of a medical image generated through data augmentation are mapped away from the center of the supersphere in the latent space. The neural network model 200 adjusts the magnitude of data augmentation in the unsupervised learning process to map the features of medical images generated through data augmentation so that they are farther away from the center of the supersphere in the latent space according to the intensity of data augmentation. can For example, the neural network model 200 may perform unsupervised learning to map features of a medical image generated through data augmentation and features of a learning image that is a target of data augmentation onto a latent space. At this time, the neural network model 200 performs an operation of mapping a feature of a medical image generated through data augmentation (or a medical image that is not suitable for the input of the interpretation model) away from the average of all features of the medical image on which data augmentation is performed. can be performed. In detail, the neural network model 200 may perform a mapping operation so that the feature of the medical image with the greater intensity of data augmentation is farther away from the mean of all features. In other words, the neural network model 200 may perform unsupervised learning to move away from the center of the supersphere in the latent space as the intensity of data augmentation for the medical image increases. Such unsupervised learning enables the neural network model 200 to effectively discriminate medical images that are not suitable for the input of the interpretation model in the latent space. In addition, the neural network model 200 calculates the initialization of a latent space in which the feature of the training image (or the medical image suitable for the input of the interpretation model), which is the object of data augmentation, is centered on the average of all features of the medical image for which data augmentation has been performed. It is possible to perform a mapping operation so that it is included in . Through these operations, the neural network model 200 includes the features of the learning image (or the medical image suitable for the input of the interpretation model) that is the object of data augmentation as much as possible in the first sphere of the latent space, and at the same time, the medical image (generated through data augmentation) Alternatively, unsupervised learning may be performed so that features of medical images that are not suitable for the input of the interpretation model are not included in the supersphere of the latent space as much as possible.

6 is a flowchart illustrating a learning process of a neural network model according to an embodiment of the present disclosure.

Referring to FIG. 6 , in step S110 , the computing device 100 according to an embodiment of the present disclosure may perform data augmentation on an input medical image for learning of a neural network model. Data augmentation may be understood as generating additional data required for learning through various image processing or conversion of medical images input for learning. Since it is not realistically easy to configure the data out of the learning distribution of the reading model with the same weight as the data included in the learning distribution of the reading model, the computing device 100 is a neural network model for estimating the input suitability for the reading model of the medical image. Data augmentation may be performed on medical images input for learning to prevent overfitting and to perform effective unsupervised learning.

In step S120, the computing device 100 may extract features of the medical image for which data augmentation has been completed through step S110 using the neural network model. The computing device 100 may individually extract features of the input images by inputting all of the data augmented medical images to the neural network model. Feature extraction of a medical image through a neural network model can be understood as a process of reducing (or compressing) the dimension of a medical image.

In step S130, the computing device 100 may perform unsupervised learning on the neural network model based on the features extracted through step S120. For example, the neural network model according to an embodiment of the present disclosure may convert features extracted through step S120 into an expression of a latent space. In this case, the neural network model may map the features extracted through step S120 to the supersphere of the latent space centered on the average of the features extracted through step S120. Specifically, the neural network model selects features of a medical image (i.e. an out-of-distribution (OOD) image that is not suitable for the input of the interpretation model) generated through data augmentation among the features extracted through step S120 in the latent space. It can be mapped to be located outside the first pitch. The neural network model may map features of a medical image (i.e., an image within a learning distribution suitable for an input of a reading model), which is a target of data augmentation, among extracted features in step S120, to be located in the supersphere of the latent space. The neural network model may be trained in a direction of minimizing a hyperphericity including most of the features of an image within a learning distribution suitable for an input of a reading model while optimizing parameters while repeatedly performing such a mapping operation.

7 and 8 are flowcharts illustrating a process of classifying medical images of a computing device according to an embodiment of the present disclosure.

Referring to FIG. 7 , in step S210, the computing device 100 according to an embodiment of the present disclosure inputs at least one medical image to a neural network model based on unsupervised learning, and inputs a model for reading a specific image. fit can be inferred. For example, the computing device 100 may estimate input suitability for a model that reads a bone image of at least one medical image using a pre-trained neural network model based on unsupervised learning through the process of FIG. 6 . can The computing device 100 may use a neural network model to estimate input suitability for a model that reads a carpal image based on a distribution between features of a medical image mapped on a latent space and features mapped through prior learning. . More specifically, the computing device 100 calculates input suitability for a model for reading a bone image based on a distance between a feature of a medical image mapped on a latent space through a neural network model and an average of features mapped through prior learning. can be estimated Accordingly, the input suitability may be a probability value indicating how far apart the feature of the medical image input to the neural network model in the latent space is from the average of pre-learned features.

In step S220, the computing device 100 may determine whether the medical image input to the neural network model corresponds to input data of a model that reads the medical image, based on the input suitability of the medical image estimated through step S210. The computing device 100 may determine input data of a model for reading a medical image based on the determination result. For example, when a model that interprets a carpal image takes a frontal carpal image as an input, a lateral carpal image, an image of another body part other than the carpal bone, such as the skull, an image including an obstruction covering a specific area of the carpal bone, or a frontal carpal bone image. An image, but of low quality, should be judged as an image that is not suitable for input of a model that reads a bone image. Therefore, for this determination, the computing device 100 compares the input suitability of the medical image estimated through step S210 with a predetermined threshold based on the input of the model for reading the bone image, and the input of the model for reading the bone image. You can also select data. The computing device 100 may use a separate neural network model for selecting input data and select input data of a model that reads the medulla bone image based on the input suitability of the medical image estimated through step S210. In other words, the computing device 100 inputs the input suitability of the medical image estimated through step S210 to a separate neural network model for selecting input data, and the medical image corresponding to the input value is input to the model that reads the hand bone image. You can finally decide whether it is suitable for your data.

Hereinafter, a process of classifying a medical image based on a model for evaluating bone age by reading a bone image according to an embodiment of the present disclosure will be described in more detail with reference to FIG. 8 .

The computing device 100 according to an embodiment of the present disclosure may obtain a medical image 50 through communication with an external system. The computing device 100 may extract features of the medical image 50 through a pretrained neural network model based on unsupervised learning, and map the extracted features to the supersphere of the latent space (S310). The supersphere of the latent space can be understood as a reference region in the latent space that encompasses features of images suitable for input of a model that evaluates bone age by reading pretrained marrow images in the unsupervised learning process of the neural network model.

The computing device 100 may calculate a distance between the center of the hypersphere of the latent space and the feature mapped to the hypersphere of the latent space through step S310 (S320). In this case, the center of the hypersphere in the latent space may correspond to the average position of features previously learned in the unsupervised learning process of the neural network model in the latent space. The computing device 100 may calculate how far apart the feature mapped from the first sphere of the latent space is through step S310 based on the average position of the previously learned features in the latent space.

The computing device 100 may estimate input suitability for a bone age evaluation model by reading the medulla bone image of the medical image 50 based on the distance calculated through step S320 (S330). The computing device 100 may calculate the input suitability score 60 of the medical image 50 based on the distance calculated through step S320. For example, as the distance in the latent space between the center of the hypersphere in the latent space and the feature of the medical image 50 mapped through step S310 is greater, the medical image 50 is a model for evaluating the bone age by reading the carpal image. It can be understood that it is not suitable for input. Therefore, as the distance between the center of the hypersphere of the latent space and the feature mapped to the hypersphere of the latent space through step S310 increases, the size of the input suitability score 60 may increase.

The computing device 100 may classify the medical image 50 by comparing the input suitability score 60 with a predetermined threshold value to finally determine the input suitability (S340). The computing device 100 may determine whether to use the medical image 50 as input data for a model that reads a bone image based on a result of comparing the input suitability score 60 with a threshold value. For example, when the input suitability score 60 is greater than or equal to a threshold value, the computing device 100 may classify the medical image 50 as input data of a model that reads a bone image (S350). When the input suitability score 60 is equal to or greater than the critical value, the computing device 100 may classify the medical image 50 as out-of-distribution (OOD) data of a model that reads a bone image (S360). At this time, it may be understood that the threshold value is predetermined based on the boundary of the second sphere of the latent space for estimating the input suitability. That is, when the feature of the medical image 50 is mapped within the boundary of the supersphere of the latent space through the neural network model, the computing device 100 interprets the input suitability score 60 as being greater than or equal to the threshold value, and the medical image 50 can be classified as input data of a model that reads a bone image. When the feature of the medical image 50 is mapped outside the perimeter of the supersphere of the latent space through the neural network model, the computing device 100 interprets the input suitability score 60 as being less than a threshold value, and converts the medical image 50 to the medulla bone. It can be classified as OOD (out-of-distribution) data of the model that reads the image.

Through the process described above, the computing device 100 can effectively classify and select images suitable for input of a model designed to read a specific medical image among medical images obtained through various paths. By filtering images to be input to the interpretation model in advance through classification and selection of such medical images, an environment in which the interpretation model can accurately receive necessary data and perform stable operations can be established.

Meanwhile, according to an embodiment of the present disclosure, a computer readable medium storing a data structure is disclosed.

Data structure can refer to the organization, management, and storage of data that enables efficient access and modification of data. Data structure may refer to the organization of data to solve a specific problem (eg, data retrieval, data storage, data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. A logical relationship between data elements may include a connection relationship between user-defined data elements. A physical relationship between data elements may include an actual relationship between data elements physically stored in a computer-readable storage medium (eg, a persistent storage device). The data structure may specifically include a set of data, a relationship between data, and a function or command applicable to the data. Through an effectively designed data structure, a computing device can perform calculations while using minimal resources of the computing device. Specifically, the computing device can increase the efficiency of operation, reading, insertion, deletion, comparison, exchange, and search through an effectively designed data structure.

The data structure can be divided into a linear data structure and a non-linear data structure according to the shape of the data structure. A linear data structure may be a structure in which only one data is connected after one data. Linear data structures may include lists, stacks, queues, and decks. A list may refer to a series of data sets in which order exists internally. The list may include a linked list. A linked list may be a data structure in which data are connected in such a way that each data is connected in a single line with a pointer. In a linked list, a pointer can contain information about connection to the next or previous data. A linked list can be expressed as a singly linked list, a doubly linked list, or a circular linked list depending on the form. A stack can be a data enumeration structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a LIFO-Last in First Out (Last in First Out) data structure. A queue is a data listing structure that allows limited access to data, and unlike a stack, it can be a data structure (FIFO-First in First Out) in which data stored later comes out later. A deck can be a data structure that can handle data from either end of the data structure.

The nonlinear data structure may be a structure in which a plurality of data are connected after one data. The non-linear data structure may include a graph data structure. A graph data structure can be defined as a vertex and an edge, and an edge can include a line connecting two different vertices. A graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in a graph data structure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. Hereinafter, a neural network is unified and described. The data structure may include a neural network. And the data structure including the neural network may be stored in a computer readable medium. The data structure including the neural network may also include preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and neural network It may include a loss function for learning of . A data structure including a neural network may include any of the components described above. That is, the data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and neural network. It may be configured to include all or any combination thereof, such as a loss function for learning of . In addition to the foregoing configurations, the data structure comprising the neural network may include any other information that determines the characteristics of the neural network. In addition, the data structure may include all types of data used or generated in the computational process of the neural network, but is not limited to the above. A computer readable medium may include a computer readable recording medium and/or a computer readable transmission medium. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes.

The data structure may include data input to the neural network. A data structure including data input to the neural network may be stored in a computer readable medium. Data input to the neural network may include training data input during a neural network learning process and/or input data input to a neural network that has been trained. Data input to the neural network may include pre-processed data and/or data subject to pre-processing. Pre-processing may include a data processing process for inputting data to a neural network. Accordingly, the data structure may include data subject to pre-processing and data generated by pre-processing. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, weights and parameters may be used in the same meaning.) Also, a data structure including weights of a neural network may be stored in a computer readable medium. A neural network may include a plurality of weights. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. A data value output from an output node may be determined based on the weight. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

As a non-limiting example, the weights may include weights that are varied during neural network training and/or weights for which neural network training has been completed. The variable weight in the neural network learning process may include a weight at the time the learning cycle starts and/or a variable weight during the learning cycle. The weights for which neural network learning has been completed may include weights for which learning cycles have been completed. Accordingly, the data structure including the weights of the neural network may include a data structure including weights that are variable during the neural network learning process and/or weights for which neural network learning is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer readable storage medium (eg, a memory or a hard disk) after going through a serialization process. Serialization can be the process of converting a data structure into a form that can be stored on the same or another computing device and later reconstructed and used. A computing device may serialize data structures to transmit and receive data over a network. The data structure including the weights of the serialized neural network may be reconstructed on the same computing device or another computing device through deserialization. The data structure including the weights of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure for increasing the efficiency of operation while minimizing the resource of the computing device (for example, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is only an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. Also, the data structure including the hyperparameters of the neural network may be stored in a computer readable medium. A hyperparameter may be a variable variable by a user. Hyperparameters include, for example, learning rate, cost function, number of learning cycle iterations, weight initialization (eg, setting the range of weight values to be targeted for weight initialization), hidden unit number (eg, the number of hidden layers and the number of nodes in the hidden layer). The foregoing data structure is only an example, and the present disclosure is not limited thereto.

9 is a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Although the present disclosure has been described above as being generally embodied by a computing device, those skilled in the art will understand that the present disclosure may be combined with computer-executable instructions and/or other program modules that may be executed on one or more computers and/or may be implemented in hardware and software. It will be appreciated that it can be implemented as a combination.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will understand that the methods of the present disclosure can be applied to single-processor or multiprocessor computer systems, minicomputers, mainframe computers as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like ( It will be appreciated that each of these may be implemented with other computer system configurations, including those that may be operative in connection with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer readable media. Computer readable media can be any medium that can be accessed by a computer, including volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. Includes removable media. By way of example, and not limitation, computer readable media may include computer readable storage media and computer readable transmission media. Computer readable storage media are volatile and nonvolatile media, transitory and non-transitory, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store desired information.

A computer readable transmission medium typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism. Including all information delivery media. The term modulated data signal means a signal that has one or more of its characteristics set or changed so as to encode information within the signal. By way of example, and not limitation, computer readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer readable transmission media.

An exemplary environment 1100 implementing various aspects of the present disclosure is shown including a computer 1102, which includes a processing unit 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 couples system components, including but not limited to system memory 1106 , to processing unit 1104 . Processing unit 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as the processing unit 1104.

System bus 1108 may be any of several types of bus structures that may additionally be interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, or EEPROM, and is a basic set of information that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

The computer 1102 may also include an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA) - the internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes—a magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to a removable diskette 1118), and an optical disk drive 1120 (e.g., a CD-ROM) for reading disc 1122 or reading from or writing to other high capacity optical media such as DVDs). The hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to the system bus 1108 by a hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to The interface 1124 for external drive implementation includes at least one or both of USB (Universal Serial Bus) and IEEE 1394 interface technologies.

These drives and their associated computer readable media provide non-volatile storage of data, data structures, computer executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art can use zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other tangible media readable by the computer, such as the like, may also be used in the exemplary operating environment and any such media may include computer executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored on the drive and RAM 1112, including an operating system 1130, one or more application programs 1132, other program modules 1134, and program data 1136. All or portions of the operating system, applications, modules and/or data may also be cached in RAM 1112. It will be appreciated that the present disclosure may be implemented in a variety of commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as a mouse 1140. Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are often connected to the processing unit 1104 through an input device interface 1142 that is connected to the system bus 1108, a parallel port, IEEE 1394 serial port, game port, USB port, IR interface, may be connected by other interfaces such as the like.

A monitor 1144 or other type of display device is also connected to the system bus 1108 through an interface such as a video adapter 1146. In addition to the monitor 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be a workstation, computing device computer, router, personal computer, handheld computer, microprocessor-based entertainment device, peer device, or other common network node, and generally includes It includes many or all of the components described for, but for simplicity, only memory storage device 1150 is shown. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and corporations and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to worldwide computer networks, such as the Internet.

When used in a LAN networking environment, computer 1102 connects to local network 1152 through wired and/or wireless communication network interfaces or adapters 1156. Adapter 1156 may facilitate wired or wireless communications to LAN 1152, which also includes a wireless access point installed therein to communicate with wireless adapter 1156. When used in a WAN networking environment, computer 1102 may include a modem 1158, be connected to a communicating computing device on WAN 1154, or establish communications over WAN 1154, such as over the Internet. have other means. A modem 1158, which may be internal or external and a wired or wireless device, is connected to the system bus 1108 through a serial port interface 1142. In a networked environment, program modules described for computer 1102, or portions thereof, may be stored on remote memory/storage device 1150. It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between computers may be used.

Computer 1102 is any wireless device or entity that is deployed and operating in wireless communication, eg, printers, scanners, desktop and/or portable computers, portable data assistants (PDAs), communication satellites, wireless detectable tags associated with It operates to communicate with arbitrary equipment or places and telephones. This includes at least Wi-Fi and Bluetooth wireless technologies. Thus, the communication may be a predefined structure as in conventional networks or simply an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet without wires. Wi-Fi is a wireless technology, such as a cell phone, that allows such devices, eg, computers, to transmit and receive data both indoors and outdoors, i.e. anywhere within coverage of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in the unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band) .

Those skilled in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields s or particles, or any combination thereof.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein are electronic hardware, (for convenience) , may be implemented by various forms of program or design code (referred to herein as software) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory devices (eg, EEPROM, cards, sticks, key drives, etc.), but are not limited thereto. Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of example approaches. Based upon design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure. The accompanying method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure, and the general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not to be limited to the embodiments presented herein, but is to be interpreted in the widest scope consistent with the principles and novel features presented herein.

Claims (13)

A method of classifying medical images performed by a computing device including at least one processor, comprising:
inputting at least one medical image to a neural network model based on unsupervised learning and estimating input suitability for a model that interprets the medical image; and
selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability;
including,
Way.
According to claim 1,
The step of estimating the suitability of an input to a model that reads the medical image,
extracting features of the at least one medical image using the neural network model;
mapping the extracted feature to a hypersphere of a latent space using the neural network model; and
estimating input suitability for a model for reading the medical image of the at least one medical image based on a feature mapped to the hypersphere of the latent space using the neural network model;
including,
Way.
According to claim 2,
The step of estimating input suitability for a model for reading the medical image of the at least one medical image based on the feature mapped to the hypersphere of the latent space,
Using the neural network model, based on the distance between the center of the hypersphere of the latent space and the feature mapped to the hypersphere of the latent space, input suitability of the at least one medical image to the model for reading the medical image is estimated. doing;
including,
Way.
According to claim 3,
The center of the supersphere of the latent space is,
Corresponding to the average of features expressed in the latent space of all medical images input to the neural network model,
Way.
According to claim 1,
The step of selecting input data of a model that reads the medical image,
selecting input data of a model that reads the medical image from the at least one medical image by comparing the estimated input suitability with a predetermined threshold;
including,
Way.
According to claim 1,
The input data of the model that reads the medical image is
An image that does not include a body part targeted by the model that reads the medical image, an image that is not captured at a position or angle that serves as a reference for reading the model that reads the medical image, or a model that reads the medical image. A medical image that is not at least one of the images containing a quality factor that interferes with the model's interpretation,
Way.
According to claim 1,
The neural network model,
Data augmentation is performed on the medical image input for learning, and pre-learning is performed through unsupervised learning based on the medical image on which the data augmentation is performed.
Way.
According to claim 1,
The neural network model,
Pre-learning through unsupervised learning that extracts the features of the medical image input for learning and maps the features of the medical image generated through data augmentation among the extracted features away from the center of the supersphere in the latent space,
Way.
According to claim 8,
The characteristics of the medical image generated through the data augmentation are:
As the magnitude of the data augmentation increases, it is mapped so as to be farther from the center of the supersphere of the latent space.
Way.
According to claim 7 or 8,
The data augmentation is
Data augmentation based on geometry transformation, data augmentation by applying contrast to medical images, data augmentation by applying noise to medical images, or artifacts or substitutions in at least part of medical images Including at least one of data augmentation applying an image,
Way.
According to claim 1,
The model for reading the medical image,
Including a model for evaluating bone age of a body part included in a medical image,
Way.
A computer program stored on a computer-readable storage medium, which, when executed in one or more processors, causes the following operations for classifying medical images to be performed, the operations comprising:
inputting at least one medical image to a neural network model based on unsupervised learning and estimating input suitability for a model that reads the medical image; and
selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability;
including,
A computer program stored on a computer-readable storage medium.
A computing device for classifying medical images,
a processor comprising at least one core;
a memory containing program codes executable by the processor; and
a network unit acquiring at least one medical image;
including,
the processor,
inputting the at least one medical image to a neural network model based on unsupervised learning, and estimating input suitability for a model that interprets the medical image;
Selecting input data of a model for reading the medical image from the at least one medical image based on the estimated input suitability;
Device.

Images