KR102336111B1 - A method and computing device for creating a neural network model that outputs brain age prediction information based on eeg data - Google Patents

Abstract

Disclosed is a method for generating a neural network model that provides brain age prediction information based on the brain wave data performed by one or more processors of a computing device for realizing the task described above. The method comprises: a step of acquiring a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users; a step of constructing learning data set based on the plurality of brain wave data and the plurality of body data; and a step of generating a brain age prediction model by performing learning for one or more network functions through the learning data set, wherein the brain age prediction model outputs the brain age prediction information by inputting the brain wave data and body data of the user as inputs. Therefore, the present invention is capable of having an effect of treating early for brain symptoms related to developmental delay.

Description

A METHOD AND COMPUTING DEVICE FOR CREATING A NEURAL NETWORK MODEL THAT OUTPUTS BRAIN AGE PREDICTION INFORMATION BASED ON EEG DATA}

The present disclosure provides analysis information based on EEG data measured from a user, and more specifically, to provide a computing device for generating a neural network algorithm for providing brain age prediction information based on EEG data.

In general, the human brain goes through a process of growth and aging as it ages.

If the brain fails to follow the normal growth process and develops late for its age, it may be linked to diseases such as brain development delay. Conversely, if the brain shows rapid aging compared to connectivity, degenerative diseases such as dementia may occur.

Meanwhile, various attempts have been made to predict the possibility of developmental disabilities in children and adolescents. For example, various attempts have been made, such as discovering a unique EEG in an autistic child by comparing the EEG between an autistic child and a normal person. Since these developmental disorders can cause difficulties in various functional areas (home, school, society, etc.) in the growth of children and adolescents, accurate prediction as early as possible is required. In particular, 3~10% of elementary school students have language development delay, which is the most common symptom of developmental delay such as autism.

In other words, measuring the brain age objectively can be a measure for accurately judging an individual's brain growth and aging progress, so that a prescription or management can be made according to the situation.

Therefore, in the art, based on EEG data, a neural network algorithm that provides information that can contribute to medical diagnosis and early treatment by identifying the brain location and frequency that causes the brain age of a subject or the cause of a developmental disorder is generated. There may be a need for a computer program to do this.

Republic of Korea Patent Publication No. 10-2017-0073557

The present disclosure has been made in response to the above-described background technology, and may provide a computing device for generating a neural network algorithm for providing brain age prediction information based on EEG data.

The problems to be solved by the present disclosure 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.

Disclosed is a method for generating a neural network model that provides brain age prediction information based on EEG data performed by one or more processors of a computing device in an embodiment of the present disclosure for solving the above-described problem. The method includes: acquiring a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users; building a learning data set based on the plurality of brain wave data and the plurality of body data; and the learning data Generating a brain age prediction model by performing learning on one or more network functions through a set, wherein the brain age prediction model outputs brain age prediction information by inputting the user's EEG data and body data as inputs can be characterized as

In an alternative embodiment, the plurality of body data is information based on prediction of brain age of each user, and includes information on age and gender corresponding to each of the plurality of users, and constructing the learning data set The step of constructing a learning input data set based on the plurality of brain wave data and constructing a learning output data set based on the plurality of body data, thereby including the learning input data set and the learning output data set building a training data set.

In an alternative embodiment, each of the plurality of EEG data is characterized in that it is acquired through each of a plurality of channels, and the step of constructing the learning data set includes: based on each age and gender of each of the plurality of users Building one or more learning input data sets by dividing the brain wave data corresponding to each of the plurality of users, and dividing each of the one or more input data sets based on the location of each of the plurality of channels corresponding to each brain wave data building one or more training input sub-data sets.

In an alternative embodiment, each of the one or more learning input sub-data sets is learning data related to each of one or more brain regions of each user, and may be characterized in that it has at least some overlapping channels.

In an alternative embodiment, the brain age prediction information includes at least one of numerical prediction information and brain age estimation factor information related to the user's brain age, and the brain age estimation factor information is used for calculating the brain age information. It may be characterized in that it includes information about the brain region and frequency based on it.

In an alternative embodiment, the method may further include performing pre-processing on the training data set, wherein performing the pre-processing may include removing features related to EEG data corresponding to a predetermined range of frequencies. can

In an alternative embodiment, performing the pre-processing comprises: performing a predetermined transformation on each of a plurality of features corresponding to each of the one or more sets of learning input data; The method may include searching for values and replacing the feature values found as outliers with a maximum feature corresponding to a maximum value among features corresponding to each of the one or more learning input data sets.

In an alternative embodiment, the step of searching for outliers based on the transformed plurality of features may include converting features having a value smaller than a first reference value or greater than a second reference value among the plurality of transformed features as outliers. determining, wherein the first reference value is determined based on a difference between a first quartile corresponding to the plurality of transformed features and a value corresponding to three times the interquartile range, and the second reference value is , may be determined based on the sum of a third quartile corresponding to the plurality of transformed features and a value corresponding to three times the interquartile range.

In an alternative embodiment, the generating the brain age prediction model comprises: utilizing each of the one or more learning input sub-data sets and a learning output data set corresponding to each of the one or more learning input sub-data sets to the one or more networks generating one or more brain age prediction sub-models by performing learning on a function, wherein the brain age prediction model is a random forest-based neural network model, and includes the one or more brain age prediction sub-models and generating the brain age prediction information based on one or more brain age prediction sub-information output through at least one of the one or more brain age prediction sub-models.

In an alternative embodiment, removing at least a portion of the training data set through cross-validation between a validation data set and a test data set, and based on a feature importance value among a plurality of features corresponding to each of the brain age prediction submodels The method may further include selecting important features by removing at least one feature.

In another embodiment of the present disclosure, a computing device for generating a neural network model providing brain age prediction information based on EEG data is disclosed. The computing device includes a processor including one or more cores, a memory for storing program codes executable in the processor, and a network unit for transmitting and receiving data to and from a user terminal, wherein the processor includes a plurality of processors corresponding to each of a plurality of users. Acquire brain wave data and a plurality of body data, build a learning data set based on the plurality of brain wave data and the plurality of body data, and perform learning on one or more network functions through the learning data set to age the brain A prediction model is generated, and the brain age prediction model may be characterized in that brain age prediction information is output by inputting the user's EEG data as an input.

In another embodiment of the present disclosure, a computer program stored in a computer-readable storage medium is disclosed. The computer program, when executed on one or more processors, causes the one or more processors to perform the following operations for generating a neural network model providing brain age prediction information based on EEG data, wherein the operations include: obtaining a plurality of brain wave data and a plurality of body data corresponding to each of the users of and generating a brain age prediction model by performing learning on a network function, wherein the brain age prediction model receives brain wave data of a user as an input and outputs brain age prediction information.

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

The present disclosure has been made in response to the above-described background technology, and may provide a computing device for generating a neural network algorithm that provides brain age prediction information based on EEG data.

In addition, it is possible to provide the effect of improving the accuracy of prediction information by identifying which features had a major influence on the prediction results through the tree-structured neural network structure and minimizing the features in stages. Accordingly, there is an expected effect of early treatment of brain symptoms related to developmental delay based on information on the location and frequency of the brain causing the cause without incurring high costs.

Effects of the present disclosure 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.

Various aspects are now described with reference to the drawings, wherein like reference numbers are used to refer to like elements collectively. In the following example, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. It will be evident, however, that such aspect(s) may be practiced without these specific details.
1 is a conceptual diagram illustrating a system in which various aspects of a computing device for generating a neural network model providing brain age prediction information based on an embodiment of the present disclosure and EEG data can be implemented.
2 is a block diagram of a computing device for generating a neural network model that provides brain age prediction information based on EEG data related to an embodiment of the present disclosure.
3 is an exemplary diagram illustrating positions of a plurality of channels related to brain wave data related to an embodiment of the present disclosure.
4 is an exemplary diagram illustrating cross-validation simulation results related to an embodiment of the present disclosure.
5 is a flowchart exemplarily illustrating a method for generating a neural network model providing brain age prediction information based on EEG data related to an embodiment of the present disclosure.
6 is a schematic diagram illustrating one or more network functions related to an embodiment of the present disclosure.

Advantages and features of the present disclosure, 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 disclosure is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only the present embodiments are provided so that the disclosure of the present disclosure is complete, and those of ordinary skill in the art to which the present disclosure belongs It is provided to fully inform those skilled in the art of the scope of the present disclosure, which 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 disclosure. 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. Accordingly, it goes without saying that the first component mentioned below may be the second component within the spirit of the present disclosure.

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

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 “unit” or “module” may be configured to reside on an addressable storage medium or to reproduce one or more processors. Thus, as an 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 as additional components and “parts” or “modules”. can be further separated.

In this specification, a computer refers to 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.

Those skilled in the art will further appreciate that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or combinations of both. It should be recognized that they can be implemented with To clearly illustrate this 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 as hardware or software depends upon 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 the present disclosure.

Hereinafter, embodiments of the present disclosure 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 conceptual diagram illustrating a system in which various aspects of a computing device for generating a neural network model providing brain age prediction information based on EEG data related to an embodiment of the present disclosure can be implemented.

A system according to embodiments of the present disclosure may include a computing device 100 , a user terminal 10 , an external server 20 , an EEG data acquisition device 30 , and a network. The components illustrated in FIG. 1 are exemplary, and additional components may be present or some of the components illustrated in FIG. 1 may be omitted. The computing device 100, the user terminal 10, the external server 20, and the EEG data acquisition device 30 according to the embodiments of the present disclosure are through a network, for the system according to the embodiments of the present disclosure. Data can be transmitted and received with each other.

Networks according to embodiments of the present disclosure include Public Switched Telephone Network (PSTN), x Digital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), Very High Speed DSL (VDSL). ), a variety of wired communication systems such as Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN) can be used.

In addition, the networks presented herein include Code Division Multi Access (CDMA), Time Division Multi Access (TDMA), Frequency Division Multi Access (FDMA), Orthogonal Frequency Division Multi Access (OFDMA), Single Carrier-FDMA (SC-FDMA) and Various wireless communication systems may be used, such as other systems.

The network according to the embodiments of the present disclosure may be configured regardless of its communication mode, such as wired and wireless, and is composed of various communication networks such as a personal area network (PAN) and a wide area network (WAN). can be In addition, the network may be a well-known World Wide Web (WWW), and may use a wireless transmission technology used for short-range communication, such as infrared (IrDA) or Bluetooth (Bluetooth). The techniques described herein may be used in the networks mentioned above, as well as in other networks.

According to an embodiment of the present disclosure, the user terminal 10 is a terminal capable of receiving brain age prediction information based on EEG data through information exchange with the computing device 100 or the EEG data acquisition device 30, and the user may mean a terminal possessed by For example, the user terminal 10 may be a terminal related to a user who wants to obtain numerical prediction information related to brain age or information related to a determining factor of the predicted brain age. Also, for example, the user terminal 10 may be a terminal related to an examiner (eg, a medical professional) that provides an analysis result related to EEG data to a user (eg, the examinee). When the user terminal 10 is a terminal related to an examiner that provides examination results to the examinee, brain age prediction information obtained from the computing device 100 is used as a medical assistant terminal for reading (or analyzing) the examination result of the examinee can be As such, the user terminal 10 is provided with a display, may receive a user's input, and provide an output of any type to the user.

The user terminal 10 may refer to any type of entity(s) in a system having a mechanism for communication with the computing device 100 . For example, the user terminal 10 is a personal computer (PC), a notebook (note book), a mobile terminal (mobile terminal), a smart phone (smart phone), a tablet PC (tablet pc), and a wearable device (wearable device) and the like, and may include all types of terminals capable of accessing a wired/wireless network. In addition, the user terminal 10 may include an arbitrary server implemented by at least one of an agent, an application programming interface (API), and a plug-in. In addition, the user terminal 10 may include an application source and/or a client application.

According to an embodiment of the present disclosure, the external server 20 may be a server that stores health checkup information or brainwave checkup information for a plurality of users. For example, the external server 20 may be at least one of a hospital server and a government server, and may be a server that stores information about EEG examination records, electronic health records, and electronic medical records for a plurality of users. For example, the EEG examination record includes information about EEG data in various frequency bands obtained through each of a plurality of channels from an area of the user's body and EEG diagnosis results corresponding to the information (eg, autism or developmental delay, etc.) may include information about Also, the external server 20 may include health checkup information related to body information of each of a plurality of users. For example, the body information of each user may include information on body age and gender.

Information stored in the external server 20 may be utilized as training data, verification data, and test data for learning the neural network in the present disclosure. That is, the external server 20 may be a server that stores data for learning the deep learning model of the present disclosure.

The computing device 100 of the present disclosure may receive EEG checkup information and health checkup information from the external server 20 , and build a learning data set based on the corresponding information. The computing device 100 may generate a brain age prediction model for calculating brain age prediction information corresponding to the user's brain wave data by performing learning on one or more network functions through the training data set.

The external server 20 is a digital device, and may be a digital device equipped with a processor, such as a laptop computer, a notebook computer, a desktop computer, a web pad, and a mobile phone, and having a computing capability with a memory. The external server 20 may be a web server that processes a service. The above-described types of servers are merely examples, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the EEG data acquisition device 30 may measure and obtain EEG data from a user. The EEG data relates to an EEG (electroencephalography) obtained from the user's body, and may be a signal obtained through a plurality of channels. For example, the EEG data acquisition device 30 may include a plurality of electrodes constituting a plurality of channels, and through the plurality of electrodes, from one or more regions of the user's body (eg, the user's scalp) to various frequency bands. Corresponding brain wave data may be acquired. As a specific example, the plurality of electrodes of the EEG data acquisition device 30 may be related to the 10-20 system. That is, 19 electrodes may be disposed on a specific region of the user's scalp. In addition, the electrodes located in each area may be divided into a plurality of frequency bands to acquire signals to observe EEG data including various frequency components.

That is, the EEG data acquisition device 30 may be a device for acquiring the user's EEG data through various combinations of channels formed by 19 electrodes and a plurality of frequency bands. The plurality of frequency bands may include, for example, delta wave, theta wave, alpha wave, beta wave 1, beta wave 2, and the like. However, the present disclosure is not limited thereto, and the plurality of frequencies divided by the EEG data acquisition device 30 in the present disclosure may further include various frequency bands.

The brain wave data acquisition device 30 may detect an electrical activity of the brain from each of a plurality of regions on the surface of the user's brain. The EEG data acquisition device 30 may include a plurality of electrodes for measuring a voltage generated based on the electrical activity of the user's brain, for example, the difference between the voltages measured through the other electrode paired with the electrode. It is possible to monitor the electrical activity of the brain. That is, the EEG data acquisition device 30 measures the voltage fluctuation due to the ion current generated through the activity of neurons inside the user's brain through a plurality of electrodes to obtain EEG data related to the electrical activity of the user's brain. can In the present disclosure, EEG data obtained through the EEG data acquisition device 30 is first EEG data obtained when a predetermined task is not performed (ie, a natural state) and a predetermined task is performed. It may include the second brain wave data obtained in the. Here, the predetermined task relates to brain wave data measured in various stimulus situations, for example, a suppression power (NoGo Suppression) task, a reflexive power (NoGo Monitoring) task, a task performance (P3b) task, a waiting condition task performance (Cue P3) task, It may include a preparation capability (CNV) task, an adaptive capability (Novelty) task, or an Ignore task.

The user's EEG data acquired through the EEG data acquisition device 30 may be transmitted to the user terminal 10 or the computing device 100 , and the EEG data transmitted to the user terminal 10 or the computing device 100 is , it can be a basis for calculating the user's brain age prediction information. The computing device 100 of the present disclosure may generate and provide brain age prediction information based on the EEG data acquired through the EEG data acquisition device 30 .

Although the brain wave data acquisition device 30 and the computing device 100 are separately represented as separate entities in FIG. 1 , the brain wave data acquisition device 30 is included in the computing device 100 according to various implementation aspects of the present disclosure. Thus, the functions of acquiring EEG data and providing analysis information corresponding to EEG data may be performed in one integrated server. In this example, when the function of the computing device 100 is integrated into the integrated server, the integrated server may perform by integrating a function of acquiring EEG data from a user and a function of generating prediction information by analyzing the obtained EEG data. .

According to an embodiment of the present disclosure, the computing device 100 may generate a brain age prediction model. In the present disclosure, the brain age prediction model may refer to an artificial neural network model that outputs brain age prediction information by inputting the user's brain wave data and body data as inputs. The computing device 100 may receive EEG checkup information or health checkup information corresponding to a plurality of users from the external server 20, and may build learning data for learning a neural network based on the information. .

More specifically, the computing device 100 may acquire a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users. The computing device 100 may receive a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users from the external server 20 . For example, the first EEG data corresponding to the first user may be data obtained by measuring an electrical signal generated when a signal is transmitted between the brain nerve cells of the first user. The EEG data may be, for example, measured based on a potential difference obtained by bringing each of a plurality of electrodes into contact with each of a plurality of regions of the user's scalp, and obtained from each electrode. The first body data corresponding to the first user may include information on the body age and gender of the first user. For example, the first body data may include information '11 years old' and 'male'. The detailed description of the aforementioned brain wave data and body data is only an example, and the present disclosure is not limited thereto.

The computing device 100 may build a learning input data set based on a plurality of brain wave data, and build a learning output data based on a plurality of body data (eg, body age). The learning data set to be constructed by the computing device 100 may include a learning input data set to which EEG data is input and a learning output data set related to a body age corresponding to each of the input data. In the present disclosure, the target value related to the output of the neural network may be the physical age of each user.

In other words, the computing device 100 takes the brain wave data of users as an input of the neural network, and the output data output in response to the input brain wave data is the learning output data (ie, body age) corresponding to the label in the corresponding brain wave data. You can learn to get closer.

The computing device 100 may generate a brain age prediction model through a learning process using the learning data set. When the generated brain age prediction model receives the user's brain wave data as an input, it may output brain age prediction information corresponding to the corresponding brain wave data. When the computing device 100 receives the user's EEG data from at least one of the user terminal 10 or the EEG data acquisition device 30, the computing device 100 processes the EEG data as an input to the brain age prediction model to provide brain age prediction information. can be printed out. Also, the computing device 100 may provide the output brain age prediction information to the user terminal 10 . The brain age prediction information of the present disclosure may include at least one of numerical prediction information related to a user's brain age and brain age estimation factor information. Here, the numerical prediction information related to the brain age may be, for example, numerical prediction information that the brain age of the first user is '8 years old'. In addition, the brain age calculation factor information may be characterized in that it includes information on brain regions and frequencies based on the calculation of the corresponding brain age information. For example, the brain age estimation factor information may include information indicating that the influence of the prefrontal region and the beta wave is large in predicting or calculating the brain age as '16'. The detailed description of the above-mentioned brain age prediction information is only an example, and the present disclosure is not limited thereto.

That is, the computing device 100 of the present disclosure builds learning data through body information including brain wave data and actual body age information of a plurality of users, and uses the constructed learning data to construct a neural network model for predicting brain age. can create In addition, by utilizing the generated neural network model, it is possible to provide brain age prediction information corresponding to EEG data. Accordingly, the user may predict the possibility of developmental delay in comparison with the actual body age through the brain age prediction information. For example, when numerical information related to brain age predicted through a neural network model is predicted to be significantly lower than actual body age, the possibility of developmental delay (ie, brain dysfunction) may be high. Accordingly, it is possible to predict or detect symptoms related to developmental delay early on the basis of EEG data measured from a user without incurring high costs. This may be to provide a measure to accurately determine the degree of brain growth and aging of infants or children and adolescents to provide a prescription or management suitable for the situation.

According to an embodiment of the present disclosure, the computing device 100 may construct one or more learning input data sets by classifying EEG data corresponding to each of the plurality of users based on the body information of each of the plurality of users. Specifically, the computing device 100 may construct one or more learning input data sets by classifying EEG data corresponding to each of the plurality of users based on the age and gender of each of the plurality of users.

In more detail, the computing device 100 may construct an infant learning input data set and an infant learning input data set by classifying the EEG data based on the age of each of the plurality of users. The computing device 100 builds an infant learning input data set through EEG data corresponding to 4 to 6 years of age, and to construct pediatric learning input data through EEG data corresponding to 6 to 19 years of age. can

Also, the computing device 100 may construct a male learning input data set and a female learning input data set by classifying brain wave data based on gender. The computing device 100 may construct a male learning input data set using brain wave data corresponding to a male gender, and may construct a female learning input data set using brain wave data corresponding to a female gender.

That is, the computing device 100 may construct one or more learning input data sets by classifying EEG data based on the age and gender of each user. The one or more learning input data sets may include, for example, a male toddler learning input data set, a female toddler learning input data set, a male toddler learning input data set, and a female toddler learning input data set.

Also, the computing device 100 may classify a plurality of learning input data sets related to learning of the neural network based on age and gender, and perform learning of each neural network through each of the plurality of learning input data sets to obtain one or more learning input data sets. A brain age prediction model can be created.

Also, the computing device 100 may receive the user's brain wave data and body data from at least one of the user terminal 10 and the brain wave data acquisition device 30 . In this case, the body data may include information on the age or gender of a user (eg, an EEG examiner). When receiving the user's brain wave data and body data corresponding to the brain wave data, the computing device 100 may determine an optimal brain age prediction model to process the brain wave data based on the body data. Specifically, the computing device 100 may receive, from the first user terminal, first EEG data related to the first user and first body data indicating that the first user corresponds to an 8-year-old male. In this case, the computing device 100 may determine an optimal brain age prediction model to process the first EEG data as an input based on the first body data.

That is, the computing device 100 may generate one or more brain age prediction models through each of one or more learning input data sets divided based on age and gender, and may generate one or more brain age prediction models using the user's body data (that is, to predict brain age). age and gender) to determine the optimal brain age prediction model. Also, it is possible to generate brain age prediction information based on EEG data by using a brain age prediction model determined through body data.

In other words, by utilizing each of one or more brain age prediction models generated through learning through learning data classified based on age and gender, it is possible to produce more optimized prediction information in response to each of a plurality of users with various body conditions. it can be possible That is, it is possible to build one or more neural network models by subdividing the learning data, and by utilizing the optimal neural network according to the user's condition, it is possible to cause improvement in output (eg, brain age prediction information) accuracy.

Also, the computing device 100 may construct one or more learning input sub-data sets by dividing each of one or more input data sets for each brain region. Specifically, the computing device 100 may construct one or more learning input sub-data sets by dividing each of one or more input data sets based on the respective positions of a plurality of channels corresponding to the EEG data.

More specifically, the EEG data of the present disclosure is obtained based on the 10-20 system, and may be obtained as 19 electrodes are disposed in a specific region of the user's scalp. In this case, the electrodes located in each region may be divided into a plurality of frequency bands to acquire signals to observe EEG data including various frequency components. That is, the EEG data corresponding to each of the plurality of users may be EEG data obtained in relation to a plurality of channels formed through 19 electrodes.

The computing device 100 divides the user's brain into one or more regions based on the positions of each channel formed by the 19 electrodes, and applies each of one or more learning input data sets to one or more learning input data sets corresponding to the divided one or more regions. Data can be divided into sub-data sets.

As a specific example, one or more learning input datasets built based on age and gender include: a male toddler learning input dataset, a female toddler learning input dataset, a male toddler learning input dataset, and a female toddler learning input dataset. may include

In this case, the computing device 100 converts each of one or more learning input data sets obtained by dividing a plurality of EEG data based on age and gender into one or more areas based on a location of each of a plurality of channels in relation to a measurement area of EEG data. It can be divided and divided into one or more learning input data sub-data sets corresponding to one or more divided regions.

The computing device 100 divides each of the learning input data sets (that is, the male infant learning input data set, the female infant learning input data set, the male infant learning input data set, and the female infant learning input data set) into four regions. The first sub-data of the learning input to the fourth sub-data of the learning input for boys corresponding to each learning input data set may be divided by division. For example, the computing device 100 divides the male infant learning input data set based on four regions of the brain, and divides the male infant learning input first sub data set, the male infant learning input second sub data set, and the male infant learning input first. 3 sub data sets and a fourth sub data set of male infant learning input can be built.

In addition, the computing device 100 performs learning on one or more network functions by utilizing each of the one or more learning input sub-data sets and the learning output data sets corresponding to each of the one or more learning input sub-data sets to predict one or more brain age You can create submodels.

That is, the brain age prediction model of the present disclosure may be configured to include one or more brain age prediction sub-models. The brain age prediction model may be constructed through a tree structure between one or more brain age prediction sub-models. For example, the brain age prediction model may include a random forest-based neural network model.

When receiving the user's brain wave data, the computing device 100 may process the corresponding brain wave data as an input of each of one or more brain age prediction sub-models. The one or more brain age prediction sub-models are neural network models learned through each learning input sub-data corresponding to each region of the brain. can In other words, each brain age prediction sub-model may output brain age prediction sub-information related to each region.

Also, the computing device 100 may generate brain age prediction information based on brain age prediction sub-information that is an output of each of one or more brain age prediction sub-models. In this case, the brain age prediction information may refer to integrated information obtained by integrating the brain age prediction sub-information output for each region through each sub-model. For example, the computing device 100 may generate brain age prediction information through an average value of brain age prediction sub-information.

Also, the computing device 100 may generate brain age prediction information based on a difference between a user's actual body age related to brain wave data to be analyzed and brain age prediction information corresponding to the analysis result. In this case, the brain age prediction information may further include brain age estimation factor information. The brain age estimation factor information may include information on brain regions and frequencies based on numerical prediction information calculation of brain age.

Additionally, the computing device 100 may provide the effect of improving the accuracy of the prediction information by identifying which features have a major influence on the prediction result through the neural network structure of the tree structure, and minimizing the features in stages. . Accordingly, by providing information related to developmental delay based on information on the location and frequency of the brain causing the cause without incurring high costs, there is an expected effect that brain symptoms due to the possibility of developmental disorders can be detected or treated early. . A detailed configuration of the computing device 100 of the present disclosure and effects according to the configuration will be described later in detail with reference to FIG. 2 .

Also, the computing device 100 may be a terminal or a server, and may include any type of device. The computing device 100 is a digital device, and may be a digital device equipped with a processor, such as a laptop computer, a notebook computer, a desktop computer, a web pad, and a mobile phone, and having a computing power having a memory. The computing device 100 may be a web server that processes a service. The above-described types of servers are merely examples, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may be a server that provides a cloud computing service, according to an embodiment of the present disclosure. More specifically, the computing device 100 is a type of Internet-based computing, and may be a server that provides a cloud computing service that processes information not with a user's computer but with another computer connected to the Internet. The cloud computing service may be a service that stores data on the Internet and allows the user to use it anytime and anywhere through Internet access without installing necessary data or programs on his/her computer. Easy to share and deliver with a click. In addition, cloud computing service not only stores data on a server on the Internet, but also allows users to perform desired tasks using the functions of applications provided on the web without installing a separate program, and multiple people can simultaneously view documents. It may be a service that allows you to work while sharing. In addition, the cloud computing service may be implemented in the form of at least one of Infrastructure as a Service (IaaS), Platform as a Service (PaaS), Software as a Service (SaaS), a virtual machine-based cloud server, and a container-based cloud server. . That is, the computing device 100 of the present disclosure may be implemented in the form of at least one of the above-described cloud computing services. The detailed description of the above-described cloud computing service is merely an example, and may include any platform for building the cloud computing environment of the present disclosure. A detailed description of a method for generating a neural network model providing brain age prediction information based on the EEG data of the present disclosure will be described later with reference to FIG. 2 below.

2 is a block diagram of a computing device for generating a neural network model that provides brain age prediction information based on EEG data related to an embodiment of the present disclosure.

As shown in FIG. 2 , the computing device 100 may include a network unit 110 , a memory 120 , and a processor 130 . Components included in the aforementioned computing device 100 are exemplary and the scope of the present disclosure is not limited to the aforementioned components. That is, additional components may be included or some of the above-described components may be omitted depending on implementation aspects for the embodiments of the present disclosure.

According to an embodiment of the present disclosure, the computing device 100 may include the user terminal 10 and the network unit 110 for transmitting and receiving data to and from the external server 20 . The network unit 110 may transmit/receive data for performing a method for generating a neural network model providing brain age prediction information according to an embodiment of the present disclosure to other computing devices, servers, and the like. That is, the network unit 110 may provide a communication function between the computing device 100 , the user terminal 10 , and the external server 20 . For example, the network unit 110 may receive an electronic health record or EEG data record for a plurality of users from a hospital server. Additionally, the network unit 110 may allow information transfer between the computing device 100 and the user terminal 10 and the external server 20 by calling a procedure to the computing device 100 .

The network unit 110 according to an embodiment of the present disclosure includes a Public Switched Telephone Network (PSTN), x Digital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), VDSL ( A variety of wired communication systems such as Very High Speed DSL), Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN) can be used.

In addition, the network unit 110 presented herein is CDMA (Code Division Multi Access), TDMA (Time Division Multi Access), FDMA (Frequency Division Multi Access), OFDMA (Orthogonal Frequency Division Multi Access), SC-FDMA ( A variety of wireless communication systems can be used, such as Single Carrier-FDMA) and other systems.

In the present disclosure, the network unit 110 may be configured regardless of its communication mode, such as wired and wireless, and may be composed of various communication networks such as a short-range network (PAN: Personal Area Network) and a local area network (WAN: Wide Area Network). can In addition, the network may be a well-known World Wide Web (WWW), and may use a wireless transmission technology used for short-range communication, such as infrared (IrDA) or Bluetooth (Bluetooth). The techniques described herein may be used in the networks mentioned above, as well as in other networks.

According to an embodiment of the present disclosure, the memory 120 may store a computer program for performing a method for generating a neural network model that provides brain age prediction information based on EEG data according to an embodiment of the present disclosure. and the stored computer program may be read and driven by the processor 130 . In addition, the memory 120 may store any type of information generated or determined by the processor 130 and any type of information received by the network unit 110 . In addition, the memory 120 may store the user's brain wave data. For example, the memory 120 may temporarily or permanently store input/output data (eg, user's brain wave data, body data, brain age prediction information corresponding to brain wave data, etc.).

According to an embodiment of the present disclosure, the memory 120 is a flash memory type, a hard disk type, a multimedia card micro type, and a card type memory (eg, a SD or XD memory, etc.), Random Access Memory (RAM), Static Random Access Memory (SRAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Programmable Read (PROM) -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 120 on the Internet. The description of the above-described memory is only an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 130 may include one or more cores, and a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a general purpose graphics processing unit (GPGPU). , data analysis such as a tensor processing unit (TPU), and a processor for deep learning.

The processor 130 may read a computer program stored in the memory 120 to perform data processing for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 130 may perform an operation for learning the neural network. The processor 130 for learning of the neural network, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating an error, updating the weight of the neural network using backpropagation calculations can be performed.

In addition, at least one of a CPU, a GPGPU, and a TPU of the processor 130 may process learning of a network function. For example, the CPU and the GPGPU can process learning of a network function and data classification using the network function. Also, in an embodiment of the present disclosure, learning of a network function and data classification using the network function may be processed by using the processors of a plurality of computing devices together. In addition, the computer program executed in the computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

In the present specification, a network function may be used interchangeably with an artificial neural network and a neural network. In the present specification, the network function may include one or more neural networks, and in this case, the output of the network function may be an ensemble of the outputs of the one or more neural networks.

In this specification, a model may include a network function. The model may include one or more network functions, in which case the output of the model may be an ensemble of outputs of the one or more network functions.

The processor 130 may read a computer program stored in the memory 120 to provide a brain age prediction model according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 130 may perform a calculation for estimating brain age prediction information based on the EEG data received from the user terminal 10 . According to an embodiment of the present disclosure, the processor 130 may perform a calculation for generating a brain age prediction model.

According to an embodiment of the present disclosure, the processor 130 may typically process the overall operation of the computing device 100 . The processor 130 processes signals, data, information, etc. input or output through the above-described components or by driving an application program stored in the memory 120 to provide or process appropriate information or functions to the user terminal. have.

According to an embodiment of the present disclosure, the processor 130 may acquire a plurality of EEG data and a plurality of body data corresponding to each of a plurality of users. Each of the plurality of EEG data corresponding to each of the plurality of users may refer to data obtained by measuring an electrical signal generated when a signal is transmitted between brain neurons of each user. The EEG data may be, for example, measured based on a voltage obtained by bringing each of a plurality of electrodes into contact with each region of the user's scalp and obtaining from each electrode. The plurality of body data is information based on prediction of the brain age of each user, and may include information on the body age and gender of each user. For example, the first body data of the first user among the plurality of body data may include information '11 years old' and 'man'. The detailed description of the aforementioned brain wave data and body data is only an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the acquisition of the plurality of brain wave data and the plurality of body data may be receiving or loading data stored in the memory 120 . Acquisition of a plurality of brain wave data and a plurality of body data is performed in another storage medium based on a wired/wireless communication means, a plurality of brain wave data related to learning of a neural network from another computing device, a separate processing module in the same computing device, and a plurality of It may be receiving or loading body data.

According to an embodiment of the present disclosure, the processor 130 may build a learning data set based on a plurality of brain wave data and a plurality of body data. In this case, the training data set is data for training the neural network, and may include a training input data set and a training output data set.

The training input data set may mean data related to one or more network functions, that is, input of a neural network, and the training output data set includes each of the output data output by the one or more network functions by taking each of the training input data sets as an input. It may mean data related to the compared correct answer.

The processor 130 may construct a learning input data set based on the plurality of brain wave data, and construct the learning output data based on the plurality of body data (eg, body age). The learning data set constructed by the processor 130 may include a learning input data set using brain wave data as an input and a learning output data set related to a body age corresponding to each of the input data. That is, in the present disclosure, the target value related to the output of the neural network may be the physical age of each user.

In other words, the processor 130 takes the brain wave data of the users as an input to the neural network, and the output data output in response to the input brain wave data is the learning output data (ie, body age) corresponding to the label in the corresponding brain wave data. You can learn to get closer.

Specifically, the processor 130 may perform learning on one or more network functions constituting the brain age prediction model using the labeled training data set. The processor 130 inputs each of the training input data sets to one or more network functions, and compares each of the output data calculated by the one or more network functions with each of the training output data sets corresponding to the labels of each of the training input data sets to obtain an error. can be derived That is, in learning of a neural network, learning input data may be input to each of the input layers of one or more network functions, and the learning output data may be compared with outputs of one or more network functions. The processor 130 may train the neural network based on an error between an operation result of one or more network functions on the learning input data and an error of the learning output data (label). The processor 130 may adjust the weights of one or more network functions in a backpropagation manner based on the error. The processor 130 may adjust the weight so that the output of the one or more network functions approaches the training output data based on an error between the operation result of the one or more network functions on the training input data and the training output data.

That is, the processor 130 may construct learning data through body information including brain wave data and actual body age information of a plurality of users, and generate a neural network model for predicting brain age by using the constructed learning data. . The brain age prediction model generated by the processor 130 may be a model characterized in that brain age prediction information is output by inputting EEG data and body data of a user (eg, an examinee). That is, the processor 130 may provide brain age prediction information corresponding to the EEG data by using the generated neural network model. In this case, the brain age prediction information may include at least one of numerical prediction information related to the user's brain age and brain age estimation factor information.

Accordingly, the user may predict the possibility of developmental delay through comparison with the actual body age through the brain age prediction information. For example, if numerical predictive information related to brain age predicted through a neural network model (ie, brain age prediction model) is predicted to be significantly lower than actual body age, the possibility of developmental delay (ie, brain dysfunction) may be high. have. Accordingly, it is possible to predict or detect symptoms related to developmental delay early on the basis of EEG data measured from a user without incurring high costs. This may be to provide a measure to accurately determine the degree of brain growth and aging of infants or children and adolescents to provide a prescription or management suitable for the situation.

According to an embodiment of the present disclosure, the processor 130 may construct one or more learning data sets by classifying EEG data corresponding to each of the plurality of users based on the body information of each of the plurality of users. Specifically, the processor 130 may construct one or more learning input data sets by classifying EEG data for a plurality of users based on each age and gender of each of the plurality of users.

More specifically, the processor 130 may construct an infant learning input data set and an infant learning input data set by classifying the EEG data based on the age of each of the plurality of users. The processor 130 may construct an infant learning input data set through EEG data corresponding to 4 to 6 years of age, and to construct pediatric learning input data through EEG data corresponding to 6 to 19 years of age. have.

In addition, the processor 130 may construct a male learning input data set and a female learning input data set by classifying the EEG data based on gender. The processor 130 may construct a male learning input data set through the brain wave data corresponding to a male gender, and construct a female learning input data set through the brain wave data corresponding to a female gender.

That is, the processor 130 may construct one or more learning input data sets by classifying the EEG data based on the age and gender of each user. The one or more learning input data sets may include, for example, a male toddler learning input data set, a female toddler learning input data set, a male toddler learning input data set, and a female toddler learning input data set. In other words, the processor 130 may generate four learning input data sets based on the age and gender of each of the plurality of users.

In addition, the processor 130 may build a plurality of learning input data sets based on age and gender, and generate one or more brain age prediction models by performing learning for each neural network through each of the plurality of learning input data sets. can do. As a specific example, the processor 130 generates a first brain age prediction model by using a male infant learning input data set, and generates a second brain age prediction model by using a female infant learning input data set, and a male child A third brain age prediction model may be generated by using the adolescent learning input data set, and a fourth brain age prediction model may be generated by using the female child and adolescent learning input data set.

In addition, the processor 130 may receive the user's brainwave data and body data from at least one of the user terminal 10 and the brainwave data acquisition device 30 . In this case, the body data may include information on the age or gender of a user (eg, an EEG examiner). When receiving the user's brain wave data and body data corresponding to the brain wave data, the processor 130 may determine an optimal brain age prediction model to process the brain wave data based on the body data. Specifically, the processor 130 may receive, from the first user terminal, first EEG data related to the first user and first body data indicating that the first user corresponds to an 8-year-old male. In this case, the processor 130 may determine an optimal brain age prediction model to process the first EEG data as an input based on the first body data.

For a specific example, the processor 130 may determine an optimal brain age prediction model for the first user to process the first EEG data based on the first body data corresponding to an 8-year-old male as the third brain age prediction model. have. The third brain age prediction model may be a neural network model trained using a male adolescent learning input data set. The processor 130 may process the first EEG data as an input to the third brain age prediction model and output brain age prediction information corresponding to the first EEG data. Specific numerical descriptions related to the above-described first user's body data and brain age prediction model are merely examples, and the present disclosure is not limited thereto.

That is, the processor 130 may generate one or more brain age prediction models through each of the one or more learning input data sets divided based on age and gender, and the body data (ie, age) of the user who wants to predict the brain age. and gender) to determine an optimal brain age prediction model. Also, it is possible to generate brain age prediction information based on EEG data by using a brain age prediction model determined through body data.

In other words, by utilizing each of one or more brain age prediction models generated through learning through learning data classified based on age and gender, in response to each of a plurality of users with various body conditions (eg, age or gender), It may be possible to calculate more optimized prediction information. That is, it is possible to build one or more neural network models by subdividing the learning data, and by utilizing the optimal neural network according to the user's condition, it is possible to cause improvement in output (eg, brain age prediction information) accuracy.

According to an embodiment of the present disclosure, the processor 130 may construct one or more learning input sub-data sets by dividing each of one or more input data sets for each brain region. Specifically, the processor 130 may construct one or more learning input sub-data sets by dividing each of one or more input data sets based on the respective positions of a plurality of channels corresponding to the EEG data.

More specifically, the EEG data of the present disclosure is obtained based on the 10-20 system, and may be obtained as 19 electrodes are disposed in a specific region of the user's scalp. In this case, the electrodes located in each region may be divided into a plurality of frequency bands to acquire signals to observe EEG data including various frequency components. That is, the EEG data corresponding to each of the plurality of users may be EEG data obtained in relation to a plurality of channels formed through 19 electrodes.

The processor 130 divides the user's brain into one or more regions based on the position of each channel formed by the 19 electrodes, and divides each one or more learning input data sets into one or more learning input data sets corresponding to the divided one or more regions. It can be divided into sub-data sets.

As a specific example, one or more learning input datasets built based on age and gender include: a male toddler learning input dataset, a female toddler learning input dataset, a male toddler learning input dataset, and a female toddler learning input dataset. may include In this case, the processor 130 converts each of the one or more learning input data sets obtained by dividing the plurality of EEG data based on age and gender into one or more areas based on the location of each of the plurality of channels in relation to the measurement area of the EEG data. It can be divided and divided into one or more learning input data sub-data sets corresponding to one or more divided regions.

The processor 130 shows each of the respective learning input data sets (i.e., the male infant learning input data set, the female infant learning input data set, the male infant learning input data set, and the female infant learning input data set), respectively, as shown in FIG. As described above, it is possible to construct the first learning input sub-data to the fourth learning input sub-data corresponding to each learning input data set by dividing the region into four regions. For a specific example, one learning input data set (eg, male infant learning input data) may be divided into four regions based on a channel formed by each electrode. For example, as shown in FIG. 3 , channels corresponding to Fp1, F7, F3, Fz, T3, C3, and Cz are divided into a first region, and corresponding to Fp2, Fz, F4, F8, Cz, C4 and T4. divided channels corresponding to T3, C3, Cz, T5, P3, Pz and 01 into a third region, and Cz, C4, T4, Pz, P4, T6 and 02 Corresponding channels may be divided into a fourth region. The processor 130 divides the male infant learning input data set based on four regions of the brain, and divides the male infant learning input first sub data set, the male infant learning input second sub data set, and the male infant learning input third sub data. Set and build a fourth sub-data set of input male infant learning. In the same way, the processor 130 may construct four learning input sub-data sets for each of four regions in correspondence with each of the female infant learning input data set, the male pediatric learning input data set, and the female pediatric learning input data set. .

That is, the processor 130 is configured for each of the four regions by one or more learning input data sets (that is, male infant learning input data, female infant learning input data set, male infant learning input data set, and female infant learning input data set). It is possible to generate each of the four training input sub-data sets based on . In this case, each of the four learning input sub-data sets may be learning data related to each of one or more brain regions of each user, and may have at least some overlapping channels. Specifically, each of the one or more regions divided by the processor 130 may have at least one overlapping channel with respect to an adjacent region.

As a specific example, as shown in FIG. 3 , in the first region and the second region, the channels corresponding to Fz and Cz are overlapped channels, and the first region and the third region are at T3, C3 and Cz. Let the corresponding channels be overlapping channels, the second and fourth regions have channels corresponding to Cz, C4 and T4 as overlapping channels, and the third and fourth regions have channels corresponding to Cz and Pz It may be characterized as an overlapping channel. In this case, the central axis Cz may be an overlapping channel corresponding to all regions.

As described above, the processor 130 may construct each of the plurality of learning input sub-data sets by dividing the EEG data for each region. In this case, the processor 130 builds each of the plurality of learning input sub-data sets to have at least some overlapping channels, thereby securing connectivity between the respective learning input sub-data sets, and integrating the information output by each sub-model later. When generating information, improved reliability and accuracy can be ensured.

According to an embodiment of the present disclosure, the processor 130 may perform preprocessing on the training data set. The pre-processing performed by the processor 130 may include pre-processing related to noise removal. Specifically, the processor 130 may perform pre-processing related to noise removal in the process of constructing EEG data as learning data for learning a neural network. For example, in order to acquire the user's brain wave data through the brain wave data acquisition device 30, a part (eg, electrode) of the device may be in contact with the actual user's scalp surface. In this case, in the detection process, motion noise according to the movement or movement of the user's eyeball, for example, an eyeball conduction or EMG signal may be included as noise in the EEG data. Accordingly, the processor 130 may perform pre-processing for removing the sleep sound. As a specific example, the processor 130 identifies and removes a horizontal eye movement component, a vertical eye movement component, a muscle movement component, or other noise components from the EEG data to perform preprocessing of extracting only EEG data effective for learning. .

Also, the preprocessing performed by the processor 130 may include preprocessing for removing data related to a specific feature. Specifically, the processor 130 may perform pre-processing of removing features related to brain wave data corresponding to frequencies in a predetermined range.

More specifically, the brain wave data of the present disclosure may include a plurality of signals divided into frequencies of various bands. In other words, the EEG data may include various frequency components. For example, frequency components of various bands may include, for example, delta wave, theta wave, alpha wave, sensor motor rhythm (SMR), first beta wave, and second beta wave. The delta wave is brain wave data having a frequency of 0.5 to 4 Hz, and may be a brain wave generated in a state of deep sleep. Theta wave is EEG data having a frequency of 4 to 7 Hz, and may be EEG data generated in a sleepy state, a distracted state, or a daydream state. Alpha wave is EEG data with a frequency of 8 to 12Hz, and may be EEG data generated when external concentration is loose in a relaxed state. SMR is EEG data with a frequency of 12 to 15 Hz, and may be EEG data generated when concentration is maintained while not moving. The first beta wave is brain wave data having a frequency of 15 to 18 Hz, and may be brain wave data generated when thinking and maintaining concentration in an active state. The second beta wave (or high beta wave) is EEG data having a frequency of 18 Hz or higher, and may be EEG data generated in a state of tension or anxiety.

Among the EEG data, the EEG related to the delta wave may be data of a frequency band that responds sensitively during the noise removal process. For example, since the delta wave has a relatively low frequency band, it may react sensitively in a preprocessing process related to noise removal (eg, the first preprocessing). Accordingly, after the pre-processing related to noise removal, the entire brain wave data may be transformed into data that is not suitable as learning data for learning the neural network by the delta wave. That is, after preprocessing for noise removal is performed on the EEG data, the output accuracy of the learned neural network may be reduced by using the EEG data as the learning data according to the influence of the delta wave-related features. Accordingly, the processor 130 may perform a pre-processing (eg, a second pre-processing) of removing features related to brain wave data corresponding to a frequency in a predetermined range. For example, the frequency in the predetermined range may mean a frequency of 0.5 to 4 Hz.

As a more specific example, the processor 130 may remove features related to a predetermined range of frequencies (eg, delta waves). The processor 130 includes all of the features related to the Delta Alpha Ratio (DAR) indicating the ratio between the delta wave and the alpha wave, and the related features composed of the delta wave and a combination of Fp1, Fp2, F7 and F8 (eg, channels related to the prefrontal region). A pretreatment to remove it can be performed.

That is, the processor 130 may perform pre-processing of removing data related to a specific feature (eg, delta wave) by removing features related to EEG data corresponding to a frequency in a predetermined range. Accordingly, since training data is built through data from which a specific feature with high sensitivity is removed in the noise removal process, the output accuracy of the neural network (ie, the brain Naietz model) of the present disclosure may be improved.

According to an embodiment of the present disclosure, the processor 130 may perform pre-processing on features related to each of a plurality of EEG data included in the training data set. In this case, the pre-processing (eg, the third pre-processing) regarding the features related to each of the EEG data may refer to a pre-processing for searching for and replacing an outlier in relation to each feature.

Specifically, the processor 130 may perform transformation on each of a plurality of features corresponding to each of one or more training input sub-data sets. Here, the predetermined transformation for the feature may mean a log transformation. Since the EEG data used as the learning data of the present disclosure has many features with a skewness value greater than 1 (meaning that the skewness value is out of 1), the processor 130 logs in response to all the features. By performing the transformation, normality of the features can be ensured.

Also, the processor 130 may search for an outlier of each of the plurality of transformed features. Because an outlier is a value that deviates a lot from the pattern, it can greatly affect the average value. For example, in [1, 2, 3, 4, 5], the mean may be 3 and the median may be 3. In addition, in the case of [1, 2, 3, 4, 100], the mean is 22, but the median may be equal to 3. That is, when such an outlier is included among the data related to learning, since it may have a great influence on the calculation of analysis or prediction results, the processor 130 performs preprocessing to search for and replace features related to the outlier. can

In more detail, the processor 130 may determine features having a value smaller than the first reference value or greater than the second reference value among the plurality of transformed features as outliers. In other words, the processor 130 may determine features other than the range of the first reference value and the second reference value as features related to the outlier. Here, the first reference value may be determined based on a difference between the first quartile (ie, Q1) of the plurality of changed features and a value corresponding to three times the interquartile range (IQR). The first quartile may mean a data value corresponding to 25% of all features. The interquartile range may be derived by the difference between the third quartile and the first quartile (ie, IQR = Q3 - Q1), which means data values corresponding to 75% of all features. As a specific example, the first reference value may be determined by satisfying an equation such as “Q1 - 3*IQR”. Also, the second reference value may be determined based on the sum of the third quartile of the transformed features and a value corresponding to three times the interquartile range. As a specific example, the second reference value may be determined by satisfying an equation such as “Q3 + 3*IQR”.

Also, the processor 130 may replace the feature values found as outliers with the maximum value feature corresponding to the maximum value among the features corresponding to each of the one or more learning input data. Here, the one or more learning input data is divided and constructed based on the age of each of the plurality of users, and may include, for example, an infant learning input data set and an infant learning input data set. For example, in relation to EEG data, the trend of each feature may be clearly displayed in proportion to age. Accordingly, the processor 130 classifies the criteria (eg, infants or children and adolescents) based on the age, and replaces the feature values found as outliers with the maximum feature corresponding to the maximum value among the features corresponding to each criterion. can do. That is, the processor 130 may perform missing value replacement for features determined as outliers among a plurality of features. This may have the effect of removing a statistical bias that occurs when simply removing features related to outliers.

According to an embodiment of the present disclosure, the processor 130 may remove at least a portion of the training data set through cross validation between the verification data set and the test data set. Each of the validation data and the test data may consist of at least a portion of the labeled training data set. For example, the verification data set may be used to determine the completion of learning based on whether the effect of learning for each epoch is greater than or less than a certain level in iterative learning of the neural network. The test data set may also be used to test the performance of one or more network functions to determine whether to activate the one or more network functions.

Removal of at least a portion of the training data set through cross-validation may mean filtering out abnormal data. As a specific example, if the output results of the neural network are different during the cross-validation process, the corresponding data may be determined as abnormal data. That is, the processor 130 may determine that learning of the neural network model is unstable because of the corresponding data, and may perform filtering on the corresponding data.

A detailed description of data removal through cross-validation performed by the processor 130 will be described below with reference to FIG. 4 . 4 is an exemplary diagram illustrating a cross-validation simulation result by way of example. A plurality of rows in the table shown in FIG. 4 may identify each of a plurality of users. For example, the first row may display data related to the first user, and the 300th row may display data related to the 300th user. In addition, a plurality of columns in the table shown in FIG. 4 may correspond to the number of times cross-validation is performed. For example, the first column may display a result of performing cross-validation once, and the 99th column may indicate a result of performing cross-validation 99 times. In addition, the record displayed corresponding to each row and column may mean a difference value between output values calculated in the case where verification data is input to the same model and when test data is input to the same model.

Specifically, the processor 130 may determine whether to remove data based on a comparison of the absolute value of the cross-validation value according to the number of times of performing each cross-validation and a predetermined threshold value. For example, as shown in FIG. 4 , the absolute value of the cross-validation value corresponding to the 8th cross-validation related to the ninth user (that is, the difference '6.4568' between the output values of the verification data and the test data) is predetermined. When the threshold value (eg, '5') is exceeded, the processor 130 may remove the learning data related to the brain wave data of the ninth user.

In addition, the processor 130 may determine whether to remove data based on a comparison between an absolute value of a difference value between cross-validation values generated for each number of times of performing cross-validation in the cross-validation process and a predetermined threshold difference value. For example, as shown in FIG. 4 , a cross-validation value corresponding to the first cross-validation related to the 19th user (ie, '2.8688') and a cross-validation value corresponding to the 10th cross-validation (ie, '0.691) '), when the absolute value (ie, '2.1778') of the difference value exceeds a predetermined threshold difference value (eg, '2'), the processor 130 removes the learning data related to the EEG data of the 19th user. can

That is, the processor 130 may remove at least a portion of the training data set for data related to abnormality (ie, data that reduces the learning efficiency of the neural network) through cross-validation. In other words, when the difference between the output data output by inputting each of the test data and the verification data as an input is relatively large or the difference between the cross-validation values output for each cross-validation cycle is relatively large, the processor 130 converts the data into the neural network. By filtering by discriminating and filtering data that lowers learning efficiency, it can be excluded from modeling. Through the above-described process, the effect of improving learning efficiency may be caused.

According to an embodiment of the present disclosure, the processor 130 may select an important feature by removing at least one feature from among a plurality of features corresponding to each of the brain age prediction submodels based on the feature importance value. Specifically, in the learning process, the brain age prediction sub-model may correspond to each of the plurality of features and output the importance value of the feature. For example, the importance value of such a feature may relate to prediction accuracy. For example, the brain age prediction sub-model may output '80' and '73' as feature importance values of the first and second features, respectively. For example, in the learning process, the processor 130 may remove features below a specific threshold (eg, 60) based on the feature importance value corresponding to each of the plurality of features, or remove features corresponding to the lower 1%. .

According to an embodiment, the number of important features finally selected for each region (eg, the first region to the fourth region) in the brain age prediction model for boys and girls is '36', '37', and '46' and '45'.

That is, the processor 130 selects only important features by stepwise minimizing the features in the learning process, so that it is easier to determine which features have a major influence. This not only improves accuracy, but also improves the explanatory power related to the output of the neural network model by extracting the features that cause problems among numerous features composed of various combinations of the position of the output EEG and the frequency derived therefrom as the number of added levels. has an improving effect.

According to an embodiment of the present disclosure, the processor 130 performs learning on one or more network functions by utilizing each of the one or more learning input sub-data sets and the learning output data sets corresponding to each of the one or more learning input sub-data sets. to generate one or more brain age prediction sub-models. For example, the processor 130 generates a brain age prediction first sub-model by using the first sub-data of the learning input related to the first region of the brain, and generates the second sub-data of the learning input related to the second region of the brain. generating a second sub-model for predicting brain age using the third sub-model of the brain age prediction, and generating a third sub-model for predicting brain age by using the third sub-data of the learning input related to the third region of the brain, and learning input related to the fourth region of the brain A fourth sub-model for predicting brain age may be generated by using the fourth sub-data. Accordingly, when the brain age prediction first sub-model receives EEG data as an input, it outputs brain age prediction sub-information related to the first region of the brain, and when the second brain age prediction sub-model receives EEG data as an input , output brain age prediction sub-information related to the second region of the brain, and when the brain age prediction third sub-model receives EEG data as input, output brain age prediction sub-information related to the third region of the brain, and When the brain age prediction fourth sub-model receives EEG data as an input, it may output brain age prediction sub-information related to the fourth region of the brain.

That is, the brain age prediction model of the present disclosure may be configured to include one or more brain age prediction sub-models. The brain age prediction model may be constructed through a tree structure between one or more brain age prediction sub-models. For example, the brain age prediction model may include a random forest-based neural network model.

When receiving the user's brain wave data, the processor 130 may process the corresponding brain wave data as input of each of one or more brain age prediction sub-models. The one or more brain age prediction sub-models are neural network models learned through each learning input sub-data corresponding to each region of the brain. can In other words, each brain age prediction sub-model may output brain age prediction sub-information related to each region.

Also, the processor 130 may generate brain age prediction information based on brain age prediction sub-information that is an output of each of one or more brain age prediction sub-models. In this case, the brain age prediction information may mean integrated information obtained by integrating brain age prediction sub-information output for each region through each model. For example, the processor 130 may generate brain age prediction information through an average value of brain age prediction sub-information. As a specific example, the brain age prediction sub-information output by the brain age prediction first sub-model with specific EEG data as input (that is, the brain age prediction sub-information output in response to the first region) is '8 years old' , the brain age prediction sub-information output by the second sub-model of the brain age prediction (that is, the sub-information output in response to the second region) is '10 years old', and the third sub-model of the brain age prediction The brain age prediction sub-information (ie, the brain age prediction sub-information output in response to the third region) is '12 years old', and the brain age prediction sub-information output by the fourth sub-model of the brain age prediction (ie, the fourth region) is When the brain age prediction sub-information output in response to) is '10 years old', the processor 130 calculates the average value of each sub-information (ie, (8+10+12+10)/4) to calculate the brain age Prediction information (ie, integrated information) may be generated as '10 years old'.

Also, the processor 130 may generate brain age prediction information based on a difference between the user's actual body age related to the brain wave data to be analyzed and the brain age prediction information corresponding to the analysis result. In this case, the brain age prediction information may further include brain age estimation factor information. The brain age estimation factor information may include information on brain regions and frequencies based on numerical prediction information calculation of brain age.

As a specific example, the actual physical age of the first user is 12, and the numerical information related to the brain age prediction output by each brain or the prediction sub-model by inputting the first EEG data of the first user is '4', ' In the case of '10', '10', and '16', the processor 130 may generate numerical prediction information of '9 years old' through the average value of each numerical information. In addition, the processor 130 develops by identifying that the brain age prediction information (ie, 9 years old) calculated based on the first EEG data of the first user is less than the actual body age of the first user (ie, 12 years old) Brain age estimation factor information related to delay can be generated. In this case, the processor 130 determines that the output result of the brain age prediction sub-model corresponding to the first region is the most different from the actual brain age, and calculates the brain age that may correspond to developmental delay in the first region. Factor information can be generated.

In addition, a region corresponding to the brain age or the brain region for which the developmental state is to be determined may be determined, and the brain age of the determined region may be determined. For example, to determine the state of the frontal lobe, the brain ages of the first and second regions are predicted, and information on the brain age corresponding to the frontal lobe is obtained by using the average of the brain ages predicted for the two regions. can do. For example, when the age of the first region is predicted to be '9 years old' and the age of the second region is predicted to be '11 years old', the brain age of the frontal lobe may be predicted to be '10 years old'. If the actual age of the subject is 12 years old, it can be seen that the development of the frontal lobe is delayed. Similarly, one or more regions corresponding to the part of the brain for which brain age or developmental status is to be determined is selected, the brain age of the selected region is predicted using EEG data of the selected region, and the predicted brain age for the selected one or more regions is selected. By combining information about brain age (eg, average), it is possible to determine the brain age or developmental state of the corresponding brain region.

Additionally, the processor 130 may provide an effect of improving the accuracy of prediction information by identifying which features have a major influence on the prediction result through the tree-structured neural network structure, and minimizing the features in stages. Accordingly, by providing information related to developmental delay based on information on the location and frequency of the brain causing the cause without incurring high costs, there is an expected effect that brain symptoms due to the possibility of developmental disorders can be detected or treated early. .

5 is a flowchart exemplarily illustrating a method for generating a neural network model providing brain age prediction information based on EEG data related to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the method may include acquiring a plurality of EEG data and a plurality of body data corresponding to each of a plurality of users ( S10 ).

According to an embodiment of the present disclosure, the method may include constructing a learning data set based on a plurality of brain wave data and a plurality of body data ( S20 ).

According to an embodiment of the present disclosure, the method may include generating a brain age prediction model by performing learning on one or more network functions through a training data set ( S30 ).

The order of the steps illustrated in FIG. 5 described above may be changed if necessary, and at least one or more steps may be omitted or added. That is, the above-described steps are merely an embodiment of the present disclosure, and the scope of the present disclosure is not limited thereto.

6 is a schematic diagram illustrating one or more network functions related to an embodiment of the present disclosure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may be composed 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 is configured to include at least one or more nodes. Nodes (or neurons) constituting neural networks may be interconnected by one or more “links”.

In the neural network, one or more nodes connected through a link may relatively form a relationship between an input node and an output node. The concepts of an input node and an output node are relative, and any node in an output node relationship with respect to one node may be in an input node relationship in a relationship with another node, and vice versa. As described above, an input node-to-output node relationship may be created around a link. One or more output nodes may be connected to one input node through a link, and vice versa.

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

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

A neural network may include one or more nodes. Some of the nodes constituting the neural network may configure one layer based on distances from the initial input node. For example, a set of nodes having a distance of n from the initial input node is You can configure n layers. The distance from the initial input node may be defined by the minimum number of links that must be passed to reach the corresponding node from the initial input node. However, the definition of such a layer is arbitrary for description, and the order of the layer in the neural network may be defined in a different way from the above. For example, a layer of nodes may be defined by a distance from the final output node.

The initial input node may mean one or more nodes to which data is directly input without going through a link in a relationship with 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. In addition, the hidden node may mean nodes constituting the neural network other than the first input node and the last output node. The neural network according to an embodiment of the present disclosure may be a neural network in which 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 progresses from the input layer to the hidden layer. can Also, in the neural network according to another embodiment of the present disclosure, the number of nodes in the input layer may be less than the number of nodes in the output layer, and the number of nodes may be reduced as the number of nodes progresses from the input layer to the hidden layer. have. In addition, the neural network according to another embodiment of the present disclosure may be 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 progresses from the input layer to the hidden layer. can The neural network according to another embodiment of the present disclosure may be a neural network in a combined form 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 be used to identify the latent structures of data. In other words, it can identify the potential structure of photos, texts, videos, voices, and music (e.g., what objects are in the photos, what the text and emotions are, what the texts and emotions 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 trust network (DBN), a Q network, a U network, a Siamese network, and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

The neural network may be trained by at least one of supervised learning, unsupervised learning, and semi-supervised learning. The training of the neural network is to minimize the error in the output. In the training of a neural network, iteratively input the training data into the neural network, calculate the output of the neural network and the target error for the training data, and calculate the error of the neural network from the output layer of the neural network to the input layer in the direction to reduce the error. It is a process of updating the weight of each node in the neural network by backpropagation in the direction. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (ie, labeled learning data), and in the case of comparative learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning related to data classification may be data in which categories are labeled in each of the learning data. The labeled training data is input to the neural network, and an error can be calculated by comparing the output (category) of the neural network with the label of the training data. As another example, in the case of comparison learning related to data classification, an error may be calculated by comparing the input training data with the neural network output. The calculated error is back propagated in the 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 change amount of the connection weight of each node to be updated may be determined according to a learning rate. The computation of the neural network on the input data and the backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of repetitions of the learning cycle of the neural network. For example, in the early stage of learning of a neural network, a high learning rate can be used to enable the neural network to quickly obtain a certain level of performance to increase efficiency, and in the late learning period, a low learning rate can be used to increase accuracy.

In the training of neural networks, in general, the training data may be a subset of real data (that is, data to be processed using the trained neural network), and thus, the error on the training data is reduced, but the error on the real data is reduced. There may be increasing learning cycles. Overfitting is a phenomenon in which errors on actual data increase by over-learning on training data as described above. For example, a phenomenon in which a neural network that has learned a cat by seeing 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 errors in machine learning algorithms. In order to prevent such overfitting, various optimization methods can be used. In order to prevent overfitting, methods such as increasing training data, regularization, or dropout in which a part of nodes in the network are omitted in the process of learning, may be applied.

Steps of a method or algorithm described in relation to an embodiment of the present disclosure may be implemented directly in hardware, as a software module executed by hardware, or 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 disclosure pertains.

Components of the present disclosure 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 disclosure 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++ , may be implemented in a programming or scripting language such as Java, assembler, or the like. Functional aspects may be implemented in an algorithm running on one or more processors.

Those of ordinary skill in the art of the present disclosure will recognize that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein include electronic hardware, (convenience For this purpose, it will be understood that it 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 upon the particular application and design constraints imposed on the overall system. A person skilled in the art of the present disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be interpreted as a departure from the scope of the present disclosure.

The various embodiments presented herein may be implemented as methods, apparatus, or articles 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 device. For example, computer-readable 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, EEPROMs, cards, sticks, key drives, etc.). Also, various storage media presented herein include one or more devices and/or other machine-readable media for storing information. The term “machine-readable medium” includes, but is not limited to, wireless channels and various other media that can store, hold, and/or convey instruction(s) and/or data.

It is to be understood that the specific order or hierarchy of steps in the presented processes is an example of exemplary approaches. Based on 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 the present disclosure. The appended 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 readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the embodiments presented herein, but is to be construed in the widest scope consistent with the principles and novel features presented herein.

Claims (12)

A method performed on one or more processors of a computing device, comprising:
obtaining, by the processor, a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users;
constructing, by the processor, a learning data set based on the plurality of brain wave data and the plurality of body data; and
generating, by the processor, learning on one or more network functions through the training data set to generate a brain age prediction model;
including,
The brain age prediction model is,
It is characterized in that it outputs brain age prediction information by inputting the user's brain wave data and body data,
The plurality of body data is
Information that is based on prediction of brain age of each user, and includes information on age and gender corresponding to each of the plurality of users,
Building the training data set comprises:
The learning comprising the learning input data set and the learning output data set, wherein the processor builds a learning input data set based on the plurality of brain wave data and a learning output data set based on the plurality of body data building a data set;
including,
Each of the plurality of EEG data,
It is characterized in that it is obtained through each of a plurality of channels,
Building the learning input data set comprises:
constructing, by the processor, one or more learning input data sets by classifying EEG data corresponding to each of the plurality of users based on the respective age and gender of each of the plurality of users; and
constructing, by the processor, one or more learning input sub-data sets by dividing each of the one or more input data sets based on the positions of each of the plurality of channels corresponding to each EEG data;
includes,
The step of generating the brain age prediction model comprises:
The processor processes each of the one or more training input sub-data sets as inputs to the one or more network functions, and outputs each of the outputs corresponding to each of the training input sub-data sets associated with the respective training input sub-data sets. generating one or more brain age prediction sub-models by comparing them with each set, deriving an error, and performing supervised learning of adjusting a connection weight of each of the one or more network functions based on the derived error;
includes,
Each of the one or more brain age prediction sub-models is associated with each of a plurality of regions related to the user's brain, and
The brain age prediction model is,
characterized in that the brain age prediction information is output by integrating one or more brain age prediction sub-information related to the output of each of the one or more brain age prediction sub-models,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
delete delete The method of claim 1,
Each of the one or more learning input sub-data sets,
Learning data relating to each of one or more brain regions of each user, characterized in that it has at least some overlapping channels,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
The method of claim 1,
The brain age prediction information,
It includes at least one of numerical prediction information and brain age calculation factor information related to the user's brain age,
The brain age calculation factor information is,
It characterized in that it includes information about the brain region and frequency based on the calculation of the brain age information,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
The method of claim 1,
performing, by the processor, pre-processing on the training data set;
further comprising,
Performing the pre-processing step,
removing, by the processor, a feature related to brain wave data corresponding to a frequency in a predetermined range;
containing,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
7. The method of claim 6,
Performing the pre-processing step,
performing, by the processor, a predetermined transformation on each of a plurality of features corresponding to each of the one or more sets of training input data;
detecting, by the processor, an outlier based on the transformed plurality of features; and
replacing, by the processor, the feature values found as outliers with a maximum feature corresponding to a maximum value among features corresponding to each of the one or more training input data sets;
containing,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
8. The method of claim 7,
The step of searching for outliers based on the transformed plurality of features includes:
determining, by the processor, features having a value less than a first reference value or greater than a second reference value among the plurality of transformed features as outliers;
including,
The first reference value is
is determined based on a difference between a first quartile corresponding to the plurality of transformed features and a value corresponding to three times the interquartile range,
The second reference value is
characterized in that it is determined based on the sum of a third quartile corresponding to the plurality of transformed features and a value corresponding to three times the interquartile range,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
The method of claim 1,
The brain age prediction model is,
A random forest-based neural network model, comprising the one or more brain age prediction sub-models,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
10. The method of claim 9,
removing, by the processor, at least a portion of the training data set through cross-validation between a validation data set and a test data set; and
selecting, by the processor, an important feature by removing at least one feature from among a plurality of features corresponding to each of the brain age prediction sub-models based on a feature importance value;
further comprising,
A method for generating a neural network model that provides brain age prediction information based on brain wave data performed by one or more processors of a computing device.
A computing device comprising:
a processor including one or more cores;
a memory storing program codes executable by the processor; and
a network unit for transmitting and receiving data to and from the user terminal;
including,
The processor is
Acquire a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users,
build a learning data set based on the plurality of brain wave data and the plurality of body data;
To generate a brain age prediction model by performing learning on one or more network functions through the training data set,
The brain age prediction model is,
It is characterized by outputting brain age prediction information by inputting the user's brain wave data,
The plurality of body data is
Information that is based on prediction of brain age of each user, and includes information on age and gender corresponding to each of the plurality of users,
Building the training data set comprises:
The learning comprising the learning input data set and the learning output data set, wherein the processor builds a learning input data set based on the plurality of brain wave data and a learning output data set based on the plurality of body data comprising building a data set;
Each of the plurality of EEG data,
It is characterized in that it is obtained through each of a plurality of channels,
Building the training input data set comprises:
The processor classifies the EEG data corresponding to each of the plurality of users based on each age and gender of each of the plurality of users to construct one or more learning input data sets, and the plurality of channels corresponding to each EEG data Splitting each of the one or more input data sets based on their respective positions to construct one or more learning input sub-data sets,
Generating the brain age prediction model is,
The processor processes each of the one or more training input sub-data sets as inputs to the one or more network functions, and outputs each of the outputs corresponding to each of the training input sub-data sets associated with the respective training input sub-data sets. generating one or more brain age prediction sub-models by comparing with each set, deriving an error, and performing teacher learning to adjust the connection weight of each of the one or more network functions based on the derived error,
Each of the one or more brain age prediction sub-models is associated with each of a plurality of regions related to the user's brain, and
The brain age prediction model is,
characterized in that the brain age prediction information is output by integrating one or more brain age prediction sub-information related to the output of each of the one or more brain age prediction sub-models,
A computing device for generating a neural network model that provides brain age prediction information based on EEG data.
A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, causes the one or more processors to generate a neural network model that provides brain age prediction information based on EEG data. to perform operations, the operations comprising:
acquiring a plurality of brain wave data and a plurality of body data corresponding to each of a plurality of users;
constructing a learning data set based on the plurality of brain wave data and the plurality of body data; and
generating a brain age prediction model by learning one or more network functions through the training data set;
including,
The brain age prediction model is,
It is characterized by outputting brain age prediction information by inputting the user's brain wave data,
The plurality of body data is
Information that is based on prediction of brain age of each user, and includes information on age and gender corresponding to each of the plurality of users,
The operation of building the training data set is,
The learning data set including the learning input data set and the learning output data set by building a learning input data set based on the plurality of brain wave data and building a learning output data set based on the plurality of body data building action;
including,
Each of the plurality of EEG data,
It is characterized in that it is obtained through each of a plurality of channels,
The operation of building the learning input data set is,
building one or more learning input data sets by classifying EEG data corresponding to each of the plurality of users based on the respective age and gender of each of the plurality of users; and
constructing one or more learning input sub-data sets by dividing each of the one or more input data sets based on the positions of each of the plurality of channels corresponding to each EEG data;
includes,
The operation of generating the brain age prediction model is,
The processor processes each of the one or more training input sub-data sets as inputs to the one or more network functions, and outputs each of the outputs corresponding to each of the training input sub-data sets associated with the respective training input sub-data sets. generating one or more brain age prediction sub-models by comparing each set to derive an error, and performing teacher learning to adjust the connection weight of each of the one or more network functions based on the derived error;
includes,
Each of the one or more brain age prediction sub-models is associated with each of a plurality of regions related to the user's brain, and
The brain age prediction model is,
characterized in that the brain age prediction information is output by integrating one or more brain age prediction sub-information related to the output of each of the one or more brain age prediction sub-models,
A computer program stored on a computer-readable storage medium.

Images