KR102423048B1 - Control method, device and program of system for determining aortic dissection based on non-contrast computed tomography images using artificial intelligence - Google Patents

Abstract

A control method of a system for judging aortic dissection based on a non-contrast computed tomography image using artificial intelligence is provided. The control method may include: acquiring, by the electronic device, at least one non-contrast computed tomography image; inputting, by the electronic device, the acquired at least one non-contrast computed tomography image into an artificial intelligence model; determining, by the electronic device, whether or not aortic dissection based on a result value output by the artificial intelligence model; includes

Description

BACKGROUND ART Control method, device and program of system for determining aortic dissection based on non-contrast computed tomography images using artificial intelligence }

The present invention relates to a control method, apparatus and program for a system for judging aortic dissection based on a non-contrast computed tomography image using artificial intelligence.

Computed tomography (computerized tomography, CT) is a non-invasive bioimaging method. In order to obtain normal and abnormal anatomical and pathological information of each part or organ of the body, computed tomography performs X-ray tomography from various angles of the human body to obtain cross-sectional (cross-sectional) images of the human body, and then computes/reconstructs these images. It is characterized in that it is implemented in three dimensions through Because computed tomography can be repeatedly performed, it is easy to observe disease progression and healing processes, and compared to MRI, the examination cost is low and the examination time is short.

However, in order to accurately diagnose a disease during computed tomography, a contrast medium is injected through a venous blood vessel. As a result of computed tomography, a region such as cancer is distinguished from normal tissue because contrast is enhanced by most contrast agents, and thus it is distinguished from normal tissue.

Aortic dissection can also be determined through the above-described contrast-enhanced computed tomography. In particular, it is difficult to make an accurate diagnosis of aortic dissection using non-contrast-enhanced computed tomography images.

However, in the case of emergency patients, the patient's condition is worsening during the time they have to wait to check the kidney function to use the contrast agent. In some cases, there is no other choice but to perform a contrast-enhanced computed tomography scan using a contrast agent.

delete

Registered Patent Publication No. 10-2058348/, 2019.12.24

An object of the present invention is to provide a control method, apparatus and program for a system for judging aortic dissection based on a non-contrast computed tomography image using artificial intelligence.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention for solving the above problems, there is provided a control method for a system for judging aortic dissection based on a non-contrast computed tomography image using artificial intelligence, in which an electronic device includes at least one non-contrast computed tomography image. acquiring an image; inputting, by the electronic device, the acquired at least one non-contrast computed tomography image into an artificial intelligence model; determining, by the electronic device, whether or not aortic dissection based on a result value output by the artificial intelligence model; includes

In this case, the artificial intelligence model may operate by at least one of an RNN algorithm using RGB numerical values of each pixel of an image and a CNN algorithm using the image itself.

In this case, the step of inputting the non-contrast computed tomography image to the AI model may include: extracting a region of interest from among the at least one computed tomography image and obtaining a boundary of the region of interest; acquiring a feature value based on the acquired region of interest; and determining whether aortic dissection includes: determining whether aortic dissection based on the feature value; may include.

In this case, the obtaining of the feature value may include: obtaining a deep feature obtained by inputting the region of interest into a DCNN (Deep Convolutional Neural Networks) model as a first feature; acquiring a texture feature obtained based on a gray level co-oc-currence matrix (GLCM) in the region of interest as a second feature; Acquiring a Fourier shape descriptor obtained based on the region of interest as a third feature; and determining whether the aortic dissection is performed by inputting the first feature into a first AI model. to obtain a first result value; obtaining a second result value by inputting the second feature into a second artificial intelligence model; obtaining a third result value by inputting the third feature into a third artificial intelligence model; and determining whether aortic dissection is present based on the first result value to the third result value.

In this case, the step of determining whether aortic dissection based on the first result value to the third result value includes ensemble of the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model to have aortic dissection. can determine whether

In this case, the control method may include: acquiring, by the electronic device, a region of interest based on the at least one non-contrast computed tomography image; dividing, by the electronic device, the at least one non-contrast computed tomography image into an ROI image and a non-ROI image; training the AI model by applying different weights to the region of interest image and the image of the non-interest region; may include.

In this case, the determining whether or not the aortic dissection includes: calculating an accuracy of whether the aortic dissection is determined based on the first to the third result values; obtaining, by the electronic device, patient information related to aortic dissection from a patient terminal; calibrating the accuracy based on the patient information; and determining whether aortic dissection is present based on the corrected accuracy. may include.

In this case, the control method may include: acquiring, by the patient terminal, information about the surrounding environment at a preset time interval; obtaining, by the patient terminal, whether an accident has occurred based on the surrounding environment information; The patient terminal, when an accident occurs, obtaining patient information about the surrounding situation; may include.

Other specific details of the invention are included in the detailed description and drawings.

According to the above-described various embodiments of the present invention, there is an effect of judging aortic dissection by a non-contrast computed tomography image.

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a system diagram according to an embodiment of the present invention.
2 is a flowchart illustrating a method for determining aortic dissection based on a non-contrast computed tomography image using artificial intelligence, according to an embodiment of the present invention.
3 and 4 are computed tomography images related to a test result of aortic dissection according to an embodiment of the present invention.
5 is an exemplary diagram for explaining an image output by inputting a non-contrast computed tomography image to an artificial intelligence model according to an embodiment of the present invention.
6 to 8 are exemplary views for explaining the operation of the artificial intelligence model according to an embodiment of the present invention.
9 is a block diagram of an apparatus according to an embodiment of the present invention.

Advantages and features of the present invention and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments allow the disclosure of the present invention to be complete, and those of ordinary skill in the art to which the present invention pertains. It is provided to fully inform those skilled in the art of the scope of the present invention, and the present invention is only defined by the scope of the claims.

The terminology used herein is for the purpose of describing the embodiments and is not intended to limit the present invention. In this specification, the singular also includes the plural, unless specifically stated otherwise in the phrase. As used herein, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other components in addition to the stated components. Like reference numerals refer to like elements throughout, and "and/or" includes each and every combination of one or more of the recited elements. Although "first", "second", etc. are used to describe various elements, these elements are not limited by these terms, of course. These terms are only used to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may be the second component within the spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used herein will have the meaning commonly understood by those of ordinary skill in the art to which this invention belongs. In addition, terms defined in a commonly used dictionary are not to be interpreted ideally or excessively unless clearly specifically defined.

As used herein, the term “unit” or “module” refers to a hardware component such as software, FPGA, or ASIC, and “unit” or “module” performs certain roles. However, “part” or “module” is not meant to be limited to software or hardware. A “part” or “module” may be configured to reside on an addressable storage medium or may be configured to reproduce one or more processors. Thus, by way of example, “part” or “module” refers to components such as software components, object-oriented software components, class components and task components, processes, functions, properties, Includes procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Components and functionality provided within “parts” or “modules” may be combined into a smaller number of components and “parts” or “modules” or additional components and “parts” or “modules”. can be further separated.

Spatially relative terms "below", "beneath", "lower", "above", "upper", etc. It can be used to easily describe the correlation between a component and other components. A spatially relative term should be understood as a term that includes different directions of components during use or operation in addition to the directions shown in the drawings. For example, when a component shown in the drawing is turned over, a component described as “beneath” or “beneath” of another component may be placed “above” of the other component. can Accordingly, the exemplary term “below” may include both directions below and above. Components may also be oriented in other orientations, and thus spatially relative terms may be interpreted according to orientation.

In this specification, a computer means all types of hardware devices including at least one processor, and may be understood as encompassing software configurations operating in the corresponding hardware device according to embodiments. For example, a computer may be understood to include, but is not limited to, smart phones, tablet PCs, desktops, notebooks, and user clients and applications running on each device.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Each step described in this specification is described as being performed by a computer, but the subject of each step is not limited thereto, and at least a portion of each step may be performed in different devices according to embodiments.

1 is a system diagram according to an embodiment of the present invention.

A system according to an embodiment of the present invention includes an electronic device 100 and a computed tomography (CT) device (hereinafter, CT device 200).

The electronic device 100 according to the present invention may include a processor, a memory, and a communication unit.

The processor controls the overall operation of the electronic device 100 .

The processor may include a RAM, a ROM, a main CPU, first to n interfaces, and a bus. In this case, the RAM, ROM, main CPU, first to n interfaces, etc. may be connected to each other through a bus.

The ROM stores an instruction set for booting the system, and the like. When a turn-on command is input and power is supplied, the main CPU copies the O/S stored in the memory to the RAM according to the command stored in the ROM, and executes the O/S to boot the system. When booting is completed, the main CPU copies various application programs stored in the memory to the RAM, and executes the application programs copied to the RAM to perform various operations.

The main CPU accesses the memory and performs booting using the O/S stored in the memory. In addition, the main CPU performs various operations using various programs, contents, data, etc. stored in the memory.

The first to n interfaces are connected to the various components described above. One of the interfaces may be a network interface connected to an external device through a network.

The memory may store an operating system (O/S) for driving the electronic device 100 . Also, the memory may store various software programs or applications for operating the electronic device 100 according to various embodiments of the present disclosure. The memory may store various information such as various data input, set, or generated during execution of a program or application.

In addition, the memory may include various software modules for operating the electronic device 100 according to various embodiments of the present disclosure, and the processor executes various software modules stored in the memory to enable electronic devices according to various embodiments of the present disclosure. An operation of the device 100 may be performed.

Also, the memory may store spatial images captured by the camera and various images received from the outside. To this end, the memory 150 may include a semiconductor memory such as a flash memory or a magnetic storage medium such as a hard disk.

The communication unit may communicate with an external device. In particular, the communication unit may include various communication chips such as a Wi-Fi chip, a Bluetooth chip, an NFC chip, a wireless communication chip, and the like. At this time, the Wi-Fi chip, the Bluetooth chip, and the NFC chip perform communication in a LAN method, a WiFi method, a Bluetooth method, and an NFC method, respectively. In the case of using a Wi-Fi chip or a Bluetooth chip, various types of connection information such as an SSID and a session key are first transmitted and received, and then various types of information can be transmitted and received after communication connection using this. The wireless communication chip refers to a chip that performs communication according to various communication standards, such as IEEE, ZigBee, 3rd Generation (3G), 3rd Generation Partnership Project (3GPP), and Long Term Evolution (LTE). In particular, the communication unit may receive various information from an external device.

However, the configuration of the electronic device 100 is not limited to the above-described configuration, and it goes without saying that various configurations may be further included. For example, it goes without saying that the electronic device 100 may further include a display and an input unit.

The display may be implemented with various types of display panels. For example, the display panel is a Liquid Crystal Display (LCD), Organic Light Emitting Diodes (OLED), Active-Matrix Organic Light-Emitting Diode (AM-OLED), Liquid Crystal on Silicon (LcoS), or Digital Light Processing (DLP). It may be implemented with various display technologies, such as. In addition, the display may be coupled to at least one of a front area, a side area, and a rear area of the electronic device 100 in the form of a flexible display.

The input unit may receive a user command. The input unit may be configured in various forms. For example, the input unit may be implemented as a touch screen in combination with a display and a touch sensing unit (not shown). However, the present invention is not limited thereto, and the input unit may include a button or may be configured as an external remote control, may be configured as a microphone for voice input, or may perform motion input in combination with a camera.

The CT device 200 is configured to transmit a non-contrast computed tomography image to the electronic device 100 .

2 is a flowchart illustrating a method for determining aortic dissection based on a non-contrast computed tomography image using artificial intelligence, according to an embodiment of the present invention.

In operation S110 , the electronic device 100 may acquire at least one non-contrast computed tomography image.

In operation S120 , the electronic device 100 may input the acquired at least one non-contrast computed tomography image to the artificial intelligence model.

At this time, the artificial intelligence model may be a CNN (Convolution Nueral Network) model, and may operate as shown in FIG. 6 , but is not limited thereto. That is, the artificial intelligence model according to the present invention may be a CNN algorithm that receives the image itself as well as an RNN algorithm that receives the RGB numerical values of each pixel of the image.

As an embodiment, the electronic device 100 may extract a region of interest from at least one computed tomography image and obtain a boundary of the region of interest.

Thereafter, the electronic device 100 may acquire a feature value based on the acquired ROI.

In operation S130, as shown in FIGS. 3 and 4 , the electronic device 100 may determine whether aortic dissection is present based on a result value output by the artificial intelligence model. 3 shows a computed tomography image in the case of aortic dissection, and FIG. 4 shows a computed tomography image in the normal case.

Specifically, the electronic device 100 may train the artificial intelligence model by inputting a non-contrast computed tomography image to the artificial intelligence model in order to detect aortic dissection. To this end, the electronic device 100 may include an image acquisition unit, a data generation unit, a data preprocessor, a learning unit, and a determination unit. However, the configuration of the electronic device 100 is not limited to those disclosed above. For example, the electronic device 100 may further include a database for storing information.

Specifically, the image acquisition unit may acquire at least one non-contrast computed tomography image.

The data generator may generate a data set. In this case, the data set may mean a grouping set of at least one non-contrast computed tomography image determined to be malicious. Alternatively, the data set may mean a grouping set of at least one non-contrast computed tomography image determined to be normal.

Alternatively, the data set may refer to a set in which information about a patient corresponding to a non-contrast computed tomography image is matched. In this case, the patient information may include various information such as the patient's gender, age, height, weight, family history, and the like. The patient information may be data input to the fully connected neural network together with the result of the convolutional neural network structure in the learning unit to be described below.

In addition, the data generator may generate a training data set and a verification data set for applying the deep learning algorithm. The data set may be generated by classifying the data set into a training data set required for artificial neural network learning and a validation data set for verifying learning progress information of the artificial neural network. In an embodiment, the data generator may randomly classify an image to be used for the training data set and an image used for the verification data set from among the non-contrast computed tomography images acquired from the CT apparatus 200 . In addition, the data generator may use the remainder of the selection of the data set for verification as the data set for training. The data set for verification may be randomly selected. A ratio of the data set for verification and the data set for training may be determined by a preset reference value. In this case, as for the preset reference value, the ratio of the verification data set may be set to 10% and the ratio of the training data set to 90%, but is not limited thereto.

The data generator may generate a data set by dividing a data set for training and a data set for verification in order to prevent an overfitting state. For example, since the training data set may be in an overfitting state due to the learning characteristics of the neural network structure, the data generator may prevent the artificial neural network from becoming overfitted by using the verification data set.

In this case, the data set for verification may be a data set that does not overlap with the data set for training. Since the data for verification is data that has not been used in the construction of the artificial neural network, it may be data that is first encountered in the artificial neural network during verification work. Therefore, the data set for verification may be a data set suitable for evaluating the performance of an artificial neural network when a new image (a new image not used for training) is input.

The preprocessor may preprocess the data set to be applicable to the deep learning algorithm. The preprocessor can preprocess the data set to improve recognition performance in deep learning algorithms and to minimize similarity with images between patients. A deep learning algorithm can consist of two parts: a structure of convolutional neural networks and a structure of fully-connected neural networks.

Meanwhile, according to an embodiment of the present application, the preprocessor may perform a preprocessing process of five steps. At least one of cropping, translation, color adjustment, and flipping of the non-contrast computed tomography image may be performed.

The preprocessor may perform a preprocessing process to correspond to a preset reference value. The preset reference value may be a value arbitrarily designated by a user. In addition, the preset reference value may be a value determined by the average value of the acquired non-contrast computed tomography images. The data set passed through the preprocessor may be provided to the learning unit.

The learning unit can build an artificial neural network through learning that takes as an input a data set that has undergone a preprocessing process and outputs an item related to a lesion classification result as an output.

According to an embodiment of the present application, the learning unit can output the lesion classification result by applying a deep learning algorithm consisting of two parts: a Convolutional Neural Networks structure and a Fully-connected Neural Networks structure. have. A fully connected deep neural network is characterized in that two-dimensional connections are made horizontally/vertically between nodes, there is no connection relationship between nodes located on the same layer, and only between nodes located in the immediately adjacent layer. is a neural network that

The learning unit can build a training model through learning using a convolutional neural network as an input to a training data set that has undergone pre-processing as an input, and a fully connected deep neural network using the output of the convolutional neural network as an input.

According to an embodiment of the present application, the convolutional neural network may extract a plurality of specific feature patterns for analyzing a non-contrast computed tomography image. In this case, the extracted specific feature pattern can be used for final classification in a fully connected deep neural network.

Convolutional Neural Networks are a type of neural network mainly used in speech recognition and image recognition. It is configured to process multidimensional array data, and is specialized for multidimensional array processing such as color images. Therefore, most of the techniques using deep learning in the image recognition field can be based on convolutional neural networks.

A convolutional neural network (CNN) can process an image by dividing it into several pieces, not just one piece of data. In this way, even if the image is distorted, the partial characteristics of the image can be extracted and correct performance can be achieved.

The convolutional neural network may have a plurality of layer structures. The elements constituting each layer may be composed of a convolutional layer, an activation function, a max pooling layer, an activation function, and a dropout layer. The convolutional layer acts as a filter called a kernel, so that partial processing of the entire image (or a new feature pattern generated) can extract a new feature pattern of the same size as the image. The convolutional layer can easily correct the values of the feature pattern through an activation function in the feature pattern. The max pooling layer can reduce the size of the image by sampling some non-contrast computed tomography images and adjusting the size. The convolutional neural network goes through the convolutional layer and the max pooling layer, so that the size of the feature pattern is reduced, but a plurality of feature patterns can be extracted by utilizing a plurality of kernels. The dropout layer may be a method that does not intentionally consider some weights for efficient training when training the weights of the convolutional neural network. On the other hand, the dropout layer may not be applied when an actual test is performed through a trained model.

A plurality of feature patterns extracted from the convolutional neural network can be transferred to the next step, a fully connected deep neural network, and used for classification. Convolutional neural networks can control the number of layers. Convolutional neural networks can build more stable models by adjusting the number of layers according to the amount of training data for model training.

In addition, the learning unit can build a judgment (training) model through learning that uses the pre-processed learning data set as an input to the convolutional neural network, and the output of the convolutional neural network and patient information as input to the fully connected deep neural network. . In other words, the learning unit may allow the image data that has undergone the preprocessing process to enter the convolutional neural network preferentially, and the result from the convolutional neural network to enter the fully connected deep neural network. In addition, the learning unit may allow arbitrarily extracted features to enter the fully connected deep neural network without going through the convolutional neural network.

The learning unit compares the error between the result obtained by applying the training data to the deep learning algorithm structure (structure formed into a fully connected deep neural network through a convolutional neural network) with the actual result, and gradually increases the weight of the neural network structure as much as the error corresponds to the error. The result can be fed back and learned through a backpropagation algorithm that changes it to . The backpropagation algorithm may be to adjust the weight from each node to the next node in order to reduce the error of the result (the difference between the actual value and the result value). The learning unit may be to derive a final judgment model by learning a neural network using a training data set and a verification data set to obtain a weight parameter.

The determination unit may preprocess the new data set and then determine the lesion through the artificial neural network. In other words, the determination unit may derive a determination on the new data by using the final determination model derived from the learning unit described above. The new data may be data including a non-contrast computed tomography image to be determined by the user. The new data set may be a data set generated by associating a new non-contrast computed tomography image with patient information. The new data set can be preprocessed in a state applicable to the deep learning algorithm through the preprocessing process of the preprocessor. Thereafter, the pre-processed new data set may be input to the learning unit, and a non-contrast computed tomography image may be determined based on the learning parameters.

By the operation of the above-described artificial intelligence model, as shown in FIG. 5 , aortic dissection may be detected based on a non-contrast computed tomography image.

According to an embodiment of the present invention, the electronic device 100 may determine whether aortic dissection is present based on the feature value.

At this time, as shown in FIG. 7 , the electronic device 100 acquires a deep feature obtained by inputting a region of interest into a DCNN (Deep Convolutional Neural Networks) model as a first feature, and selects the region of interest as a first feature. A texture feature obtained based on a gray level co-oc-currence matrix (GLCM) may be obtained as a second feature, and a Fourier shape descriptor obtained based on a region of interest may be obtained as a third feature.

Thereafter, the electronic device 100 inputs the first feature to the first artificial intelligence model to obtain a first result value w1, and inputs the second feature to the second artificial intelligence model to obtain a second result value w2 ), and input the third feature to the third artificial intelligence model to obtain a third result value w3.

Thereafter, the electronic device 100 may determine whether aortic dissection is present based on the first to third result values. For example, as shown in FIG. 7 , it is possible to determine whether benign or malignant by summing the first to third result values.

Meanwhile, as another embodiment of the present invention, the first AI model, the second AI model, and the third AI model may be ensembled to determine whether aortic dissection is present.

Specifically, as shown in FIG. 8 , the first to third artificial intelligence models may include a conv layer, a ReLU layer, and a Softmax layer. The ensemble model according to FIG. 8 may be a basic ensemble model (Simple Ensemble) that ensembles final output values of the first artificial intelligence model to the third artificial intelligence model.

However, according to various embodiments of the present invention, the ensemble model does not transfer the output value of the ReLU layer to the next layer, but transfers the ensemble value of the output value of each ReLU layer to the next layer. Of course, it can be an ensemble) model. In this case, there is an effect of saving runtime memory.

As another embodiment, it is similar to the Stepwise Ensemble model, but when it is not appropriate to ensemble all layers as needed, a Stepwise Partial Ensemble model may be used.

For example, the electronic device 100 may perform a step-wise ensemble of the first artificial intelligence model and the second artificial intelligence model, and may output the result value of the third artificial intelligence model as it is.

As another embodiment, when the performance of the electronic device 100 is limited, the electronic device 100 may apply a cascade partial ensemble model. For example, the electronic device 100 performs a step-wise ensemble of the first artificial intelligence model and the second artificial intelligence model, and the ensembled output value is an input value of one artificial intelligence model (eg, the fourth artificial intelligence model). and the final output values of the fourth artificial intelligence model and the third artificial intelligence model can be ensembled. Such a cascade ensemble model can save memory by reducing the number of artificial intelligence models as the steps are repeated. That is, when there are a plurality of artificial intelligence models predicted to have similar results, a cascade ensemble model may be used.

As another embodiment, of course, the electronic device 100 may apply a stochastic partial ensemble model for deriving a stable result value. The probabilistic ensemble model ensembles at least two artificial intelligence models among a plurality of artificial intelligence models, but one output value may be ensembled with a plurality of artificial intelligence models, and this ensemble is determined probabilistically. For example, in the probabilistic ensemble model, the first artificial intelligence model and the second artificial intelligence model are ensembled, the first artificial intelligence model and the third artificial intelligence model are ensembled, and the second artificial intelligence model and the third artificial intelligence model are ensembled. and the three ensembled output values can be used as input values for two AI models.

Meanwhile, according to various embodiments of the present disclosure, the electronic device 100 may acquire a region of interest based on at least one non-contrast computed tomography image.

Thereafter, the electronic device 100 divides at least one non-contrast computed tomography image into an ROI image and a non-ROI image, and applies different weights to the ROI image and the non-ROI image to generate an artificial intelligence model. Of course, it can be learned.

Specifically, the electronic device 100 may apply the weight so that the weight applied to the ROI is higher than the weight applied to the non-interest region. The electronic device 100 may train the artificial intelligence model according to the applied weight. Thereafter, when a non-contrast computed tomography image is input, the electronic device 100 may acquire a region of interest and a region of non-interest based on the learned artificial intelligence model.

Meanwhile, according to various embodiments of the present disclosure, the electronic device 100 may calculate the accuracy indicating the accuracy of the determination result regarding whether the aortic dissection is determined based on the first result value to the third result value.

Thereafter, the electronic device 100 may acquire patient information related to aortic dissection from the patient terminal.

Thereafter, the electronic device 100 may correct accuracy based on patient information.

Thereafter, the electronic device 100 may determine whether aortic dissection is present based on the corrected accuracy.

For example, even if the accuracy obtained based on the first to third results is low, the patient is a hypertensive patient or is pregnant, Marfan syndrome, bicuspid aortic valve, Ehlers-Danlos syndrome, Turner syndrome, etc. In the case of including congenital factors of the aortic dissection, it is determined that the probability of occurrence of aortic dissection is high and the accuracy can be corrected.

Alternatively, it goes without saying that the accuracy of the electronic device 100 may be corrected based on the patient's age or gender. For example, the electronic device 100 may correct the accuracy based on information on the status of occurrence of aortic dissection for each age of the patient.

On the other hand, a case of suspected aortic dissection may occur due to the occurrence of an accident. In this case, if the situation information at the time of the accident can be obtained, it is possible to more accurately grasp the condition of the patient. Accordingly, it goes without saying that the electronic device 100 may determine whether aortic dissection is present by collecting information on the accident location from the patient terminal.

Specifically, the patient terminal may acquire surrounding environment information at preset time intervals. In an embodiment, the patient terminal may acquire information about the surrounding audio source by turning on a microphone built into the patient terminal at a preset time interval (eg, 5 minutes).

Meanwhile, according to various embodiments of the present disclosure, the patient terminal may turn on a microphone built into the patient terminal when the location information received through the GPS of the patient terminal is rapidly changed. For example, when the change in the position of the patient terminal rapidly decreases at a preset speed or more, or vice versa, the patient terminal may turn on the microphone built into the patient terminal.

Thereafter, the patient terminal may acquire whether an accident has occurred based on the surrounding environment information.

Specifically, the patient terminal may determine the size of the acquired audio source. For example, when the acquired audio source is greater than or equal to a preset decibel, the patient terminal may detect that an accident has occurred. Alternatively, the patient terminal may determine that an accident has occurred when the decibel of the acquired audio source increases at a predetermined period.

Alternatively, the patient terminal may extract a human voice from among the acquired audio sources and analyze the extracted voice. When the extracted voice is a voice related to a rescue, an accident may occur in the patient terminal. A method of determining that the extracted voice is a voice related to a structure may be performed by a method of determining whether a pre-stored voice, for example, a voice such as “Help” or “No one” is included. Alternatively, the patient terminal may detect whether an accident has occurred based on natural language processing (NLP) technology.

In another embodiment, when the patient terminal receives a specific sound through the microphone, it may determine that an accident has occurred. For example, when the patient terminal acquires the sound of an ambulance siren as an audio source, the patient terminal may determine that an accident has occurred.

Meanwhile, according to an embodiment of the present invention, the patient terminal may acquire the first audio source information for a preset first time (eg, 1 second).

The patient terminal analyzes the first audio source, and when the analysis result satisfies the first criterion, the patient terminal may acquire second audio source information for a second preset time (eg, 5 seconds).

The patient terminal analyzes the second audio source, and when the analysis result satisfies the second criterion, the patient terminal may acquire the second audio source information for a third preset time period (eg, 5 minutes).

This is to prevent the problem that the battery of the patient terminal is quickly consumed when the audio source is periodically received for a long time, the first audio source is acquired for a short period of time, and if there is a possibility of an accident, the first audio source is longer than the preset first time. This is to receive the audio source for a second preset time, which is a long time, and to receive the audio source for a third preset time when the probability of an accident occurring is high.

In this case, the first criterion may be a case in which the decibel of the first audio source is equal to or greater than a preset level. That is, since the preset first time is a very short time, size information of the audio source may be acquired as the first reference.

Meanwhile, the preset second criterion may be whether the second audio source includes specific audio (eg, ambulance siren sound).

Meanwhile, the third criterion may be whether the third audio signal includes a specific voice (eg, a voice such as “Help” or “No one”).

That is, the first criterion and the third criterion may mean the maximum amount of information that can be analyzed during the recording time (the first time to the third time preset). At the first short time, only the size of the audio source may be determined, at the second time longer than the first time, the type of the audio source may be determined, and at the long third time, the voice included in the audio source may be analyzed.

delete

delete

9 is a block diagram of an apparatus according to an embodiment of the present invention.

The processor 102 may include one or more cores (not shown) and a graphic processing unit (not shown) and/or a connection path (eg, a bus, etc.) for transmitting and receiving signals to and from other components. .

The processor 102 according to an embodiment performs the method described with respect to FIGS. 1 to 8 by executing one or more instructions stored in the memory 104 .

For example, the processor 102 obtains new training data by executing one or more instructions stored in the memory, and performs a test on the acquired new training data by using the learned model, and the test result, labeling Extracting the first learning data in which the obtained information is obtained with an accuracy greater than or equal to a predetermined first reference value, deleting the extracted first learning data from the new learning data, and removing the new learning data from which the extracted learning data is deleted It is possible to retrain the learned model using

On the other hand, the processor 102 is a RAM (Random Access Memory, not shown) and ROM (Read-Only Memory: ROM) for temporarily and / or permanently storing a signal (or, data) processed inside the processor 102. , not shown) may be further included. In addition, the processor 102 may be implemented in the form of a system on chip (SoC) including at least one of a graphic processing unit, a RAM, and a ROM.

The memory 104 may store programs (one or more instructions) for processing and controlling the processor 102 . Programs stored in the memory 104 may be divided into a plurality of modules according to functions.

The steps of the method or algorithm described in relation to the embodiment of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or implemented by a combination thereof. A software module may include random access memory (RAM), read only memory (ROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, hard disk, removable disk, CD-ROM, or It may reside in any type of computer-readable recording medium well known in the art to which the present invention pertains.

The components of the present invention may be implemented as a program (or application) to be executed in combination with a computer, which is hardware, and stored in a medium. Components of the present invention may be implemented as software programming or software components, and similarly, embodiments may include various algorithms implemented as data structures, processes, routines, or combinations of other programming constructs, including C, C++ , Java, assembler, etc. may be implemented in a programming or scripting language. Functional aspects may be implemented in an algorithm running on one or more processors.

As mentioned above, although embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art to which the present invention pertains can realize that the present invention can be embodied in other specific forms without changing its technical spirit or essential features. you will be able to understand Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive.

100: electronic device
200: CT device

Claims (10)

In the control method of a system for judging aortic dissection based on a non-contrast computed tomography image using artificial intelligence,
acquiring, by the electronic device, at least one non-contrast computed tomography image;
inputting, by the electronic device, the acquired at least one non-contrast computed tomography image into an artificial intelligence model;
determining, by the electronic device, whether or not aortic dissection based on a result value output by the artificial intelligence model; including,
The step of inputting the non-contrast computed tomography image to the artificial intelligence model comprises:
extracting a region of interest from the at least one computed tomography image and obtaining a boundary of the region of interest;
Including; acquiring a feature value based on the acquired region of interest;
The step of determining whether the aortic dissection is,
determining whether aortic dissection is present based on the feature value; including,
The step of obtaining the feature value is
obtaining a deep feature obtained by inputting the region of interest into a DCNN (Deep Convolutional Neural Networks) model as a first feature;
acquiring a texture feature obtained based on a gray level co-oc-currence matrix (GLCM) in the region of interest as a second feature;
Including; acquiring a Fourier shape descriptor obtained based on the region of interest as a third feature;
The step of determining whether the aortic dissection is,
obtaining a first result value by inputting the first feature into a first artificial intelligence model;
obtaining a second result value by inputting the second feature into a second artificial intelligence model;
obtaining a third result value by inputting the third feature into a third artificial intelligence model; and
Determining whether or not aortic dissection based on the first result value to the third result value;
The step of determining whether aortic dissection based on the first result value to the third result value comprises:
determining whether or not aortic dissection by ensemble of the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model;
The step of determining whether the aortic dissection is,
calculating an accuracy indicating accuracy of a determination result regarding whether aortic dissection is determined based on the first result value to the third result value;
obtaining, by the electronic device, patient information related to aortic dissection from a patient terminal;
calibrating the accuracy based on the patient information; and
determining whether aortic dissection is present based on the corrected accuracy; including,
The control method is
acquiring, by the patient terminal, surrounding environment information at preset time intervals;
Including, by the patient terminal, whether an accident has occurred based on the surrounding environment information;
The step of obtaining the surrounding environment information includes:
Including, by the patient terminal, turning on the microphone to obtain information about the surrounding audio source;
The step of obtaining the surrounding audio source information includes:
obtaining, by the patient terminal, first audio source information for a first preset time;
The patient terminal analyzes the first audio source, and when the analysis result satisfies the first criterion, acquiring second audio source information for a second preset time period;
The first criterion is
When the decibel of the first audio source is greater than or equal to a preset size,
The preset second time is
longer than the preset first time,
The step of obtaining whether the accident has occurred is
determining, by the patient terminal, that an accident has occurred when it is determined that the voice extracted from the acquired audio source is a voice related to the rescue, or when the siren sound of an ambulance is acquired from the audio source; Way.
According to claim 1,
The artificial intelligence model is a control method, characterized in that it operates by at least one of a CNN algorithm and an RNN algorithm. delete delete delete According to claim 1,
The control method is
acquiring, by the electronic device, a region of interest based on the at least one non-contrast computed tomography image;
dividing, by the electronic device, the at least one non-contrast computed tomography image into an ROI image and a non-ROI image;
training the AI model by applying different weights to the region of interest image and the image of the non-interest region; A control method comprising a. delete delete a memory storing one or more instructions; and
a processor executing the one or more instructions stored in the memory;
The processor by executing the one or more instructions,
An apparatus for performing the method of claim 1 . A computer program stored in a computer-readable recording medium in combination with a computer, which is hardware, to perform the method of claim 1.

Images