KR102344532B1 - Explainable artificial intelligence system for diagnosis of mental diseases and method for diagnosing of mental diseases - Google Patents

Abstract

The present invention relates to an explainable artificial intelligence system and comprises: a communication unit for receiving the patient's brain waves; and a processor for preprocessing the measured brain waves through noise removal and epoching processing, extracting at least one first brain wave characteristic from the preprocessed brain waves, determining at least one second brain wave characteristic necessary for diagnosing the patient's mental diseases from among the at least one first brain wave characteristic and the importance of each of the at least one second brain wave characteristic using a machine learning model trained for diagnosing mental diseases, generating a decision-making structure for diagnosing the patient's mental diseases, diagnosing the mental disease of the patient by substituting the at least one second brain wave characteristic and the importance into the decision-making structure, and visualizing and providing the decision-making structure, which is explanatory information for explaining the diagnosis result and the basis for the diagnosis. An object of the present invention is to provide an explainable artificial intelligence system for diagnosing mental diseases that can explain the decision-making process leading to the diagnosis of mental diseases. The present patent is the result of the support from Gyeonggi-do Economic Science Promotion Agency with the financial resources of 2020 Gyeonggi-do global startup enterprise support project.

Description

EXPLAINABLE ARTIFICIAL INTELLIGENCE SYSTEM FOR DIAGNOSIS OF MENTAL DISEASES AND METHOD FOR DIAGNOSING OF MENTAL DISEASES

The present invention relates to an explainable artificial intelligence system for the diagnosis of mental disorders.

Currently, for the diagnosis of mental illness, the patient's mental illness is identified through a self-response questionnaire or a psychiatrist diagnoses the patient according to the DSM-5 standard. For example, self-response questionnaires such as BAI (The Beck Anxiety Inventory) and STAI (The State-Trait Anxiety Inventory) were used for anxiety disorders, and self-response questionnaires such as BDI (The Beck Depression Inventory) for depression. was used to identify the psychiatric illness suffered.

However, in the case of the self-response questionnaire, there is a problem that the results may be biased by the subjectivity of the respondent. For example, even with the same symptoms, different diagnosis results may be obtained because the evaluation criteria are different for each individual. There is a high problem with this.

In addition, when a medical person such as a psychiatrist makes a diagnosis, there is a possibility that an erroneous diagnosis result may be obtained due to the medical practitioner's incomplete judgment.

The introduction of an artificial intelligence system to prevent incomplete judgment by medical personnel and provide assistance to medical personnel in diagnosis is being actively discussed. However, most AI systems only provide the results of decision making and have a black box structure that cannot explain the process or rationale leading to the decision.

Artificial intelligence systems, which have a structure that makes it impossible to explain the process leading to decision-making, are difficult to actively use in the medical field, where even the slightest mistake can lead to fatal results, no matter how high the diagnostic accuracy.

"This patent is the result of receiving support from the Gyeonggi Provincial Economic Science Promotion Agency with the financial resources of the 2020 Gyeonggi-do Global Startup Enterprise Support Project"

Patent Publication No. 10-2019-0069047 (2019.06.19.)

An object of the present invention for solving the above problems is to provide an explanatory artificial intelligence system for diagnosing a mental disease that can explain a decision-making process leading to a diagnosis of a mental disease.

The present inventors explainable artificial intelligence system for solving the above problems is a communication unit for receiving the patient's brain waves; and pre-processing the measured EEG through noise removal and epoching processing, extracting at least one first EEG characteristic from the pre-processed EEG, and using a machine learning model learned for diagnosing mental disorders, the At least one second EEG characteristic necessary for diagnosing the patient's mental illness among at least one first EEG characteristic and the importance of each of the at least one second EEG characteristic are determined, and decision making for diagnosing the patient's mental illness Generating a structure, substituting the at least one second EEG characteristic and the importance to the decision-making structure to diagnose the patient's mental illness, and to explain the diagnosis result and the basis for the diagnosis. A processor that visualizes and provides a decision-making structure; includes, and the machine learning model may use an EEG for each age group and an EEG characteristic for each channel included in the EEG for each age as learning data for diagnosing a mental disease.

In addition, the processor provides a description and importance of the at least one second EEG for understanding the decision-making structure together with the decision-making structure, and the decision-making structure may be a tree structure.

In addition, the description of the at least one second EEG characteristic includes a channel name necessary for diagnosing the patient's mental illness, an EEG type in the channel, EEG power, and a connection diagram between channels, and the diagnosis result may be information on at least one mental disease that the patient suffers from.

Also, in each step of the decision-making structure, the processor compares any one of the at least one second EEG characteristic with a threshold value, and, according to the comparison result, compares the next second EEG characteristic with a threshold value A step may be determined.

In addition, the processor may diagnose the mental disorder of the patient based on the importance and the at least one second EEG characteristic in the lowest stage of the decision-making structure.

Also, the machine learning model may extract the at least one second EEG characteristic using a variational autoencoder.

In addition, the learning data may further include feedback information of medical staff on the diagnosis of the mental illness.

A method of diagnosing a mental disease through an explainable artificial intelligence of the present invention for solving the above-described problem, the method comprising: receiving a patient's brain waves; pre-processing the measured EEG through noise removal and epoching; extracting at least one first EEG characteristic from the preprocessed EEG; Using a machine learning model learned for diagnosing mental illness, at least one second EEG characteristic necessary for diagnosing the patient's mental illness among the at least one first EEG characteristic and the importance of each of the at least one second EEG characteristic determining; generating a decision-making structure for diagnosing the patient's mental illness; diagnosing the mental disorder of the patient by substituting the at least one second EEG characteristic and the importance into the decision-making structure; and visualizing and providing the decision-making structure as explanatory information for explaining the diagnosis result and the basis for the diagnosis, wherein the machine learning model is included in the age-specific EEG and the age-specific EEG for diagnosing mental disorders The EEG characteristics of each channel can be used as learning data.

According to the embodiments disclosed in the present invention, the explainable artificial intelligence system can be used in the medical field because it is possible to explain the decision-making process leading to the diagnosis of mental illness to the user.

1 is a schematic diagram for collecting learning data according to an embodiment of the present invention.
2 is a block diagram illustrating an XAI system according to an embodiment of the present invention.
3 is a schematic diagram illustrating diagnosing a mental disorder using the XAI system according to an embodiment of the present invention.
4A to 4E are schematic diagrams illustrating diagnosing mental disorders using the XAI system according to an embodiment of the present invention.
5 is an exemplary diagram of a decision-making structure provided by the XAI system according to an embodiment of the present invention.
6 is a flowchart illustrating diagnosing a mental disorder using the XAI system according to an embodiment of the present invention.

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

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

The word "exemplary" is used herein in the sense of "used as an illustration or illustration." Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

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

Also, in this specification, all “units” may be controlled by at least one processor, and at least one processor may perform operations performed by the “units” of the present disclosure.

Embodiments of the present specification may be described in terms of a function or a block performing a function. Blocks, which may be referred to as 'parts' or 'modules', etc. in the present disclosure include logic gates, integrated circuits, microprocessors, microcontrollers, memories, passive electronic components, active electronic components, optical components, hardwired circuits, and the like. It may be physically implemented by analog or digital circuitry, such as, and optionally driven by firmware and software.

Embodiments of the present specification may be implemented using at least one software program running on at least one hardware device and may perform a network management function to control an element.

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

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

1 is a schematic diagram for collecting learning data according to an embodiment of the present invention.

The XAI (eXplainable AI, hereinafter XAI system) system 100, which is an explanatory artificial intelligence system of the present invention, can diagnose a patient's mental illness, and provide a diagnosis result and a diagnostic basis together. The XAI system 100 may communicate with at least one server 110a, 110b, 110c, 110d, ... 110n in order to collect learning data for learning. In this case, the at least one server 110a, 110b, 110c, 110d, ... 110n may include a cloud server such as Amazon Web Services (AWS) or MS Azure. In addition, the at least one server (110a, 110b, 110c, 110d, ... 110n) may include a calculation server for calculating the EEG signal in order to analyze the EEG signal to determine the mental illness suffering. Accordingly, the XAI system 100 may obtain analysis data of various EEG signals and mental disorders corresponding to specific EEG signals from the at least one server 110a, 110b, 110c, 110d, ... 110n.

The XAI system 100 may communicate with at least one server 110a, 110b, 110c, 110d, ... 110n using the network 120, and the network 120 may be a wired, wireless communication link or optical fiber cable. It may include a connection part (not shown) such as In addition, the network 120 may be implemented as a variety of different networks, such as an intranet, a local area network (LAN), or a wide area network (WAN).

The XAI system 100 of the present invention can diagnose a patient's mental illness through a decision-making structure and brain waves using a machine learning model learned for diagnosing mental disorders, such as deep learning.

Deep learning can refer to an artificial neural network-based machine learning method that simulates human biological neurons and allows machines to learn.

As an example of a machine learning model, a deep neural network (DNN) may include a system or network that builds one or more layers in one or more computers and performs a judgment based on a plurality of data. .

The deep neural network may be implemented as a set of layers including a convolutional pooling layer, a locally-connected layer, and a fully-connected layer.

The convolutional pooling layer or local connection layer may be configured to extract features of brain waves. The fully connected layer may determine correlations between brain wave features.

As another example, the overall structure of the deep neural network of the present invention may be formed in a form in which a local access layer is connected to a convolutional pooling layer, and a fully connected layer is formed in the local access layer. The deep neural network may include various judgment criteria (ie, parameters), and may add new judgment criteria (ie, parameters) through input EEG analysis.

The deep neural network according to embodiments of the present invention is a structure called a convolutional neural network suitable for EEG analysis, and a feature extraction layer that learns by itself the feature with the greatest discriminative power from the given image data. ) and a prediction layer that learns a predictive model to obtain the highest predictive performance based on the extracted features may have an integrated structure.

The feature extraction layer spatially integrates a convolution layer, which creates a feature map by applying a plurality of filters, and a feature map to extract features that are invariant to changes in position or rotation. It may be formed in a structure in which the pooling layer is alternately repeated several times. Through this, various level features can be extracted from low-level features such as points, lines, and planes to complex and meaningful high-level features.

The convolutional layer obtains a feature map by taking a nonlinear activation function on the dot product of the filter and the local receptive field for each patch of input, and compares it with other network structures. Therefore, CNNs are characterized by using filters with sparse connectivity and shared weights. Such a connection structure reduces the number of parameters to be learned, makes learning through the backpropagation algorithm efficient, and consequently improves prediction performance.

The features finally extracted through iteration of the convolutional layer and the integration layer are fully connected layer (Fully-connected Layer) by classification models such as multi-layer perception (MLP) or support vector machine (SVM). ) and can be used for classification model training and prediction.

Also, according to an embodiment of the present invention, training data for machine learning may be generated based on the U-Net-dhSgement model. Here, the U-Net-dhSgement model is based on an end-to-end fully connected convolutional network (FCN), and contracts an expansive path and a symmetric path. It could be a model that created a U-shaped architecture with a skip connection for each level by setting it to .

According to an embodiment of the present invention, the machine learning model can learn for diagnosing mental disorders using learning data including at least one of age-specific brain waves, analysis data of the brain waves, and mental disorders corresponding to the characteristics of the brain waves. . The machine learning model may use, as learning data, EEG characteristics for each age, EEG characteristics included in the preprocessed EEG for each age group, and mental disorders corresponding to EEG characteristics for each channel among the EEG characteristics. Specifically, the machine learning model may use, as learning data, the mental illness of a patient finally derived as a diagnosis result by substituting the EEG characteristics for each channel into a decision-making structure to be described later.

In addition, the machine learning model can use the feedback information of medical staff on the diagnosis of mental illness as learning data. When the XAI system 100 provides the diagnosis result and the decision-making structure that is the basis for diagnosis as visual information, the medical staff may decide whether to trust the information according to their own judgment. When it is determined that the diagnosis result or decision-making structure of the XAI system 100 is incorrect, the medical staff may correct the wrong part and provide it to the XAI system 100 as feedback information, and the XAI system 100 provides feedback information of the medical staff can be used as training data.

2 is a block diagram illustrating an XAI system 200 according to an embodiment of the present invention. The XAI system 200 of FIG. 2 may correspond to the XAI system 100 of FIG. 1 .

According to an embodiment of the present invention, the XAI system 200 may include a communication unit 210 , a memory 220 , and a processor 230 . The components shown in FIG. 2 are not essential for implementing the XAI system 200, so the XAI system 200 described herein may have more or fewer components than those listed above. have. For example, among the components, the communication unit 210 may include one or more modules that enable wireless communication with an external device (not shown) or an external server.

EEG refers to recording the microscopic electrical activity of a group of brain cells by attaching electrodes to the scalp, amplifying them by the EEG system, and recording potential on the vertical axis and time on the horizontal axis. In other words, EEG measurement is to measure electrical activity occurring in the cerebral cortex. EEG changes spatio-temporally according to brain activity, state at the time of measurement, and brain function. can have In addition, brain waves are classified into delta waves (δ waves, Delta waves), theta waves (θ waves, Theta waves), alpha waves (α waves, Alpha waves), beta waves (β waves, Beta waves) according to the frequency range. Characteristics of each EEG may exist for each frequency.

According to one embodiment of the present invention, EEG (Electroencephalogram) refers to electroencephalography, and electrical recording signals recorded by inducing electric potential fluctuations occurring in the brain of humans or animals, or brain currents caused by them, on the scalp. means MEG (Magnetoencephalogram) means an EEG, and it means a signal recorded by measuring the microscopic biomagnetism generated from the electrical activity of brain nerve cells with a SQUID sensor or the like. ECoG (Electrocorticogram) means cortical conduction, and it means an electrical recording signal recorded by directly measuring potential fluctuations occurring in the cerebrum or brain current caused by it by implanting electrodes from the surface of the cerebral cortex. NIRS (Near-infrared spectroscopy) refers to a near-infrared spectrometer, and the NIRS EEG signal that can be used in the present invention refers to a signal that measures and records a difference between a low-level light wave reflected by the brain. In the present specification, EEG, MEG, ECoG, etc. are exemplified, but the EEG signal is not limited to the specific type of EEG signal, and all signals that can be measured in the human head are generated from the human brain. can be referred to

According to an embodiment of the present invention, the patient's EEG may be measured using a portable EEG measuring device such as Emotiv, OpenBci, or NeuroSky. In this case, the patient may measure the EEG using a portable EEG measurement device owned by the patient and then transmit the EEG to the hospital, or may receive EEG measurement equipment from the hospital to measure his/her EEG and then transmit it to the hospital. Through this, the patient can measure his or her EEG in order to be diagnosed with a mental illness at a desired time or place without going to the hospital in a non-face-to-face manner. The measured EEG may be received through the communication unit 210 of the XAI system 200 .

In addition, if the patient visits the hospital directly, the patient's EEG signal can be measured by a head cap manufactured by Biosemi (registered trademark) equipped with 64 electrodes. For example, in a hospital, medical staff may use Electrical Geodesics Inc. Geodesic™ electroencephalography (not shown) sold by (EGI) (registered trademark), or other types of electroencephalography devices, such as those sold by Compumedics NeuroScan (registered trademark), which usually calculates between 16 to 256 electrodes (not shown) may be used to measure the patient's brain waves. If the patient visits the hospital directly, the XAI system 200 may receive the EEG of the patient measured in the hospital through the communication unit 210 .

The communication unit 210 of the present invention may communicate with various types of external devices according to various types of communication methods. The communication unit 210 may include at least one of a Wi-Fi chip, a Bluetooth chip, a wireless communication chip, and an NFC chip.

According to the mobile communication technology of the present specification, technical standards or communication methods (eg, Global System for Mobile communication (GSM), Code Division Multi Access (CDMA), Code Division Multi Access 2000 (CDMA2000), Enhanced EV-DO (EV-DO) Voice-Data Optimized or Enhanced Voice-Data Only), Wideband CDMA (WCDMA), High Speed Downlink Packet Access (HSDPA), High Speed Uplink Packet Access (HSUPA), Long Term Evolution (LTE), Long Term Evolution (LTE-A) -Advanced), etc.), transmits/receives radio signals to and from at least one of a base station, an external terminal, and an external server on a mobile communication network.

In addition, as the wireless Internet technology of the present specification, for example, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband) ), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed Downlink Packet Access), HSUPA (High Speed Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), and the like.

In addition, the short-range communication technology of the present specification, Bluetooth (Bluetooth™), RFID (Radio Frequency Identification), infrared communication (Infrared Data Association; IrDA), UWB (Ultra Wideband), ZigBee, NFC (Near Field Communication), Wi- A technology supporting short-range communication by using at least one of Wireless-Fidelity (Fi), Wi-Fi Direct, and Wireless Universal Serial Bus (USB) technology may be included.

According to an embodiment of the present invention, the memory 220 is a local storage medium supporting various functions of the XAI system 200 . The memory 220 may store analysis data of the EEG signal including the EEG signal received by the communication unit 210 , the type of mental illness corresponding to the specific EEG signal, and the like. In addition, the memory 220 may store a plurality of application programs (application programs or applications) driven in the XAI system 200 , data for the operation of the XAI system 200 , and instructions. At least some of these application programs may be downloaded from an external server through wireless communication. The application program may be stored in the memory 220 , installed on the XAI system 200 , and driven to perform an operation (or function) of the XAI system 200 by the processor 230 .

In addition, the memory 220 of the present invention should retain data even when the power supplied to the XAI system 200 is cut off, and may be provided as a writable non-volatile memory (Writable Rom) to reflect changes. That is, the memory 220 may be provided with either a flash memory, an EPROM, or an EEPROM. In the present invention, it is described that all instruction information is stored in one memory 220 for convenience of description, but the present invention is not limited thereto, and the XAI system 200 may include a plurality of memories.

The processor 230 generally controls the overall operation of the XAI system 200 in addition to the operation related to the application program. The processor 230 may provide or process appropriate information or functions to the user by processing signals, data, information, etc. input or output through the above-described components or by driving an application program stored in the memory 220 .

In addition, the processor 230 may control at least some of the components of FIG. 2 in order to drive an application program stored in the memory 220 . Furthermore, the processor 230 may operate by combining at least two or more of the components included in the XAI system 200 to drive the application program.

Hereinafter, an operation of the processor 230 of the XAI system 200 will be described with reference to FIGS. 2 to 6 .

According to an embodiment of the present invention, when a patient's brain wave is received through the communication unit 210 , the processor 230 may perform pre-processing.

In general, when measuring EEG, noise may occur due to heartbeat or body movements. Accordingly, the processor 230 may perform a preprocessing process of removing unnecessary high-frequency components and low-frequency components for diagnosing mental disorders through brain wave analysis and removing artifacts due to movement.

According to an embodiment of the present invention, the processor 230 may pre-process the received EEG through noise removal or filtering. Specifically, the processor 230 may use an Independent Component Analysis (Independent Component Analysis) or a Principal Component Analysis (Principle Component Analysis) method for removing EMG (Electromyography) or EOG (electrooculogram) noise to remove noise. can

In addition, the processor 230 may include any one of a low-pass filter, a high-pass filter, a band-pass filter, and a notch filter to remove noise. can be used to remove noise. For example, in addition to general noise signals along the general transmission path (wired and wireless channels), other biosignals other than EEG signals such as EMG (Electromyogram) and EOG (Electrooculogram, safety level) are also signals of interest. Since it is not, the processor 230 may treat it as noise and remove it through filtering or the like.

In addition, according to an embodiment of the present invention, epoching refers to cutting noise-removed EEG data into a specific section for signal processing, and epoching is performed from tens of milliseconds to seconds ( second) can be used.

According to an embodiment of the present invention, the processor 230 may extract at least one EEG characteristic from the preprocessed EEG. In this case, the processor 230 may extract power for each frequency through spectral density analysis, and may extract quantitative EEG characteristics using linear, nonlinear, or complex network analysis.

Specifically, the processor 230 uses a technique such as Fourier Transform, Partial Directed Coherence (PDC), Direct Transfer Function (DTF), Independent Component Analysis (ICA), Principle Component Analysis (PCA), or Common Spatial Pattern (CSP). In this way, EEG characteristics can be extracted. In addition, the EEG characteristics include event related potential (ERP), steady state visually evoked potential (SSVEP), event related synchronization (ERS), and event related desynchronization (ERD: event related desynchronization), a default mode network (DMN), or a combination thereof.

According to an embodiment of the present invention, the first EEG extracted through the processor 230 is EEG (eg, γ wave, α wave, β wave, δ wave, θ wave, etc.) power, frequency and channel for each frequency of the patient. It may include interconnectivity and the like. Here, the channel may include a plurality of points on the scalp where the EEG of the patient is measured. In addition, the inter-channel connection diagram is a phase locking value (PLV) between brainwave signals, a correlation coefficient, an interference coefficient, a Granger's Causality Index, and a Partial Directed Coherence (PDC). ), Directed Transfer Function (DTF), mutual information, transfer entropy, and synchronization similarity.

When the at least one first EEG characteristic is extracted, the processor 230 uses the machine learning model learned for diagnosing a mental disorder, at least one second EEG characteristic necessary for diagnosing a patient's mental disorder and the at least one 2 Mental disorders can be diagnosed through the importance of each EEG characteristic.

In addition, according to an embodiment of the present invention, the processor 230 generates the decision-making structure as explanatory information for explaining the diagnosis result and the basis for the diagnosis, and visualizes the generated decision-making structure to provide visual information to the medical staff. can provide

3 is a schematic diagram illustrating diagnosing a mental disorder using the XAI system according to an embodiment of the present invention.

First, from the patient's brainwave, the first brainwaves of F1 to Fn may be extracted through preprocessing and feature extraction [S310].

Next, by inputting the first EEG signal as an input, the XAI system 200 can diagnose the patient's mental illness through the explanatory XAI process process [S320]. Here, the explainable XAI process process may refer to a series of processes for diagnosing mental disorders using the machine learning model of the present specification.

When the diagnosis of the patient's mental illness is completed, the XAI system 200 may visualize the diagnosis result, the EEG signal related to the diagnosis of mental illness, and the importance information on the EEG signal related to the diagnosis and provide it to the medical staff [S330].

Hereinafter, a process for diagnosing a mental disorder will be described in more detail with reference to FIGS. 4A to 4E .

4A to 4E are schematic diagrams illustrating diagnosing mental disorders using the XAI system according to an embodiment of the present invention, and FIG. 5 is an exemplary diagram of a decision-making structure provided by the XAI system.

First, referring to FIG. 4A , the patient's EEG may be measured using the EEG measuring device [S410].

The patient's brain wave may be measured by an electroencephalogram measuring device such as Emotiv, OpenBci, NeuroSky, or GeodesicTM and received by the communication unit 210 of the XAI system 200 through a network.

Referring to FIG. 4B , the received EEG 410 may be pre-processed through noise removal and epoch processing [S420].

Specifically, the received EEG 410 is subjected to noise removal and epoch processing by the EEG signal preprocessor 420 that removes noise using at least one of a low-pass filter, a high-pass filter, a band-pass filter, and a notch filter. can be pre-treated. In this case, when the XAI system 200 includes the EEG signal preprocessor 420 , the processor 230 may control the EEG signal preprocessor 420 to perform preprocessing. Alternatively, when the EEG signal preprocessor 420 is an external device of the XAI system 200 , the XAI system 200 receives the EEG 430 preprocessed through the EEG signal preprocessor 420 through the communication unit 210 . can do.

When the preprocessing is completed, the processor 230 may extract at least one first EEG characteristic from the preprocessed EEG 430 [S430].

Specifically, the processor 230 may extract power for each frequency through spectral density analysis, and may extract quantitative EEG characteristics using linear, nonlinear, or complex network analysis.

Next, the processor 230 may diagnose a mental illness by inputting the extracted at least one first EEG into the machine learning model learned for diagnosing mental illness [S440].

In this case, at least one of the patient's age, the patient's mental state identified through the self-response questionnaire, and the patient's health state information identified by the medical staff may be additionally input to the machine learning model.

According to an embodiment of the present invention, the processor 230 is configured to provide at least one second EEG characteristic necessary for diagnosing the patient's mental illness among at least one first EEG characteristic and the importance of each of the at least one second EEG characteristic. can be decided

Here, the second EEG characteristic may include the patient's frequency-specific EEG (eg, γ-wave, α-wave, β-wave, δ-wave, θ-wave, etc.) power, frequency, and connectivity between channels.

The processor 230 is a machine learned through an auto-encoder or a Variational Autoencoder (VAE) to determine at least one second EEG characteristic necessary for diagnosing a patient's mental illness among at least one first EEG characteristic. Learning models are available.

According to an embodiment of the present invention, the processor 230 may determine not only the at least one second EEG characteristic, but also the importance of each of the second EEG characteristics using a machine learning model.

When the second EEG characteristic and importance are determined, the processor 230 may diagnose the patient's mental illness through the decision-making structure [S450].

In the present invention, the decision-making structure may be a series of processes that the processor 230 sequentially considers, compares, and calculates to diagnose a patient's mental illness.

According to an embodiment of the present invention, the decision-making structure may be a tree structure in which the mental disorder of the patient is diagnosed based on the importance and the at least one second EEG characteristic in the lowest stage.

The processor 230 may generate a decision-making structure for diagnosing a patient's mental illness, and may diagnose the patient's mental illness by substituting at least one second EEG characteristic and importance into the decision-making structure.

Referring to FIG. 5 , the processor 230 may visualize and provide a diagnosis result 530 and the decision-making structure 510 , which is explanatory information for explaining the basis for diagnosis.

As shown in FIG. 5 , the processor 230 may compare any one of at least one second EEG characteristic with a threshold value in each step of the decision-making structure 510 . In addition, according to the comparison result, the next step of comparing the second EEG and the threshold value may be determined. For example, the processor 230 may compare XAI-F1, which is one of the second EEG characteristics, with the first specific threshold in the uppermost stage of the decision-making structure 510 . In this case, according to the comparison result, XAI-F2, which is one of the second EEG characteristics in the second step of the decision-making structure 510, and the second specific threshold value to be compared in size may be different.

In the present invention, the second EEG characteristics compared in the comparison step of the same turn may be the same or different. In addition, the grain boundary values compared with the respective second EEG characteristics may be the same or different. For example, the second EEG characteristic to be compared in the second comparison step of the decision-making structure 510 is the same as XAI-F2, and the threshold value compared to XAI-F2 may vary depending on the comparison result of the upper step, which is the first comparison step. have. In addition, in the third comparison step of the decision-making structure 510 , the second EEG characteristic to be compared may be different from XAI-F3 or XAI-F5, and the threshold value may also be different from 70 or 50.

According to an embodiment of the present invention, the processor 230 may diagnose the mental illness of the patient based on the importance and at least one second EEG characteristic in the lowest stage of the decision-making structure 510 . Specifically, the processor 230 may assign the determined importance to at least one second EEG characteristic and each of the second EEG characteristic, and diagnose the patient's mental illness through the machine learning model.

When the diagnosis of the mental disorder is completed, the processor 230 may visualize and provide the diagnosis result 530 and the decision-making structure 510 used for diagnosis. Accordingly, the medical staff is included in the decision-making structure 510 from the provided visual information to determine at least one second EEG characteristic used for diagnosing a patient's mental illness, a threshold used as a comparison value, whether it is normal, and the mental illness the patient is suffering from. Able to know. In addition, the processor 230 may also provide information on the calculation formula used in the lowest stage when diagnosing a mental illness. The processor 230 may diagnose the mental illness the patient suffers from by calculating the second EEG characteristic and importance through the above calculation formula.

According to an embodiment of the present invention, the diagnosis result 530 may be information on at least one mental disorder. Most mental disorders are not independent, so patients may have multiple mental disorders. Accordingly, as a result of diagnosing a mental illness, when the patient suffers from a plurality of mental diseases, the processor 230 may provide information on at least one mental disease that the patient suffers as a relative value.

When the diagnosis of mental illness is completed, the diagnosis result 530 and the decision-making structure 510 used for diagnosis may be visualized and provided. Therefore, the medical staff can have credibility in the XAI system 200 by checking information related to the diagnosis basis of the XAI system 200 as well as the mental illness the patient is suffering from through the provided visual information. Therefore, since the XAI system 200 according to an embodiment of the present invention can explain the decision-making process leading to the diagnosis of mental illness to the medical staff, it can be actively used in the medical field by increasing the user's confidence in the diagnosis result. have.

According to an embodiment of the present invention, the processor 230 may provide descriptive information 520 for understanding the decision-making structure 510 together with the decision-making structure 510 .

Here, the description information 520 may include a description and importance of at least one second EEG characteristic. The description of the at least one second EEG characteristic may include at least one of a channel name used for diagnosing a patient's mental illness, an EEG type in the channel, EEG power, and a connection diagram between channels.

For example, referring to FIG. 5 , the processor 230 through the explanatory information 520, the second EEG characteristics (XAI-F1, XAI-F2, ..., XAI-Fm), the name of the channel corresponding to each of the second EEG characteristics, the type of EEG in the channel used, and the EEG power and/or the connection between channels may be provided along with their respective importance for the second EEG characteristic. .

6 is a flowchart illustrating diagnosing a mental disorder using the XAI system according to an embodiment of the present invention.

Each step of the method for diagnosing mental illness of the present invention may be performed by various types of electronic devices such as the XAI system 200 including the communication unit 210 , the memory 220 , and the processor 230 .

Hereinafter, the method for diagnosing mental disorders of the present invention by the processor 230 will be described in detail with reference to FIG. 6 .

The embodiments described for the XAI system 200 are applicable to at least some or all of the methods for diagnosing mental disorders, and on the contrary, the embodiments described for the methods for diagnosing mental disorders are at least to the embodiments for the XAI system 200 . Some or all of them are applicable. In addition, the method for diagnosing mental illness according to the disclosed embodiments is not limited to being performed by the XAI system 200 disclosed herein, and may be performed by various types of electronic devices.

First, the XAI system 200 may receive the patient's brain waves through the communication unit 210 [S610].

Next, the processor 230 may pre-process the EEG measured through noise removal and epoching [S620].

According to an embodiment of the present invention, the pre-processing step ( S620 ) may not be performed by the processor 230 . Specifically, the EEG preprocessing is performed through an external device (not shown) or an external server (not shown), and the XAI system 200 may receive only the preprocessed EEG signal. In this case, the step of receiving the patient's brain wave (S610) and the pre-processing step (S620) may be omitted.

Next, the processor 230 may extract at least one first EEG characteristic from the preprocessed EEG [S630].

Next, the processor 230 uses the machine learning model learned for diagnosing a mental illness, at least one second EEG characteristic necessary for diagnosing a patient's mental illness among at least one first EEG characteristic, and the at least one second EEG characteristic. 2 It is possible to determine the importance of each EEG characteristic [S640].

Next, the processor 230 may generate a decision-making structure for diagnosing a patient's mental illness [S650].

Next, the processor 230 may diagnose the patient's mental illness by substituting at least one second EEG characteristic and importance into the decision-making structure [S660].

Finally, the processor 230 may provide a visualization result and a decision-making structure [S670].

Here, the visual information provided by visualization may be provided to an external display (not shown) having a display function through the communication unit 210 . Alternatively, when the XAI system 200 includes a display (not shown), the XAI system 200 may provide the diagnosis result, the decision-making structure 510 and the explanatory information 520 to the medical staff through the display. .

Various embodiments of the present invention include one or more instructions stored in a storage medium (eg, memory) readable by a machine (eg, XAI system 200 or computer). It can be implemented as software. For example, the processor (eg, the processor 230 ) of the device may call at least one of the one or more instructions stored from the storage medium and execute it. This makes it possible for the device to be operated to perform at least one function according to the called at least one instruction. The one or more instructions may include code generated by a compiler or code executable by an interpreter. The device-readable storage medium may be provided in the form of a non-transitory storage medium. Here, the 'non-transitory storage medium' is a tangible device and only means that it does not include a signal (eg, electromagnetic wave), and this term means that data is semi-permanently stored in the storage medium. and temporary storage. For example, the 'non-transitory storage medium' may include a buffer in which data is temporarily stored.

According to one embodiment, the method according to various embodiments disclosed herein may be provided as included in a computer program product. Computer program products may be traded between sellers and buyers as commodities. The computer program product is distributed in the form of a machine-readable storage medium (eg compact disc read only memory (CD-ROM)), or through an application store (eg Play Store™) or on two user devices. It can be distributed (eg downloaded or uploaded) directly or online between devices (eg smartphones). In the case of online distribution, at least a portion of the computer program product (eg, a downloadable app) is stored at least on a machine-readable storage medium, such as a memory of a manufacturer's server, a server of an application store, or a relay server. It may be temporarily stored or temporarily created. As mentioned above, although embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art to which the present invention pertains know that the present invention may be embodied in other specific forms without changing the technical spirit or essential features thereof. you will be able to understand Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive.

100, 200: XAI system 210: communication department
220: memory 230: processor

Claims (10)

Communication unit for receiving the patient's brain waves; and
Pre-processing the received EEG through noise removal and epoching processing,
extracting at least one first EEG characteristic from the preprocessed EEG,
Using a machine learning model learned for diagnosing mental illness, at least one second EEG characteristic necessary for diagnosing the patient's mental illness among the at least one first EEG characteristic and the importance of each of the at least one second EEG characteristic to decide,
generating a decision-making structure for diagnosing the patient's mental illness,
diagnosing the mental illness of the patient by substituting the at least one second EEG characteristic and the importance into the decision-making structure;
a processor that visualizes and provides the decision-making structure as explanatory information for explaining the diagnosis result and the basis for the diagnosis;
The diagnosis result is information about at least one mental disorder suffering from the patient,
The description of the at least one second EEG characteristic includes a channel name required for diagnosing the patient's mental illness, an EEG type and EEG power in the channel, and a connection diagram between channels,
The decision-making structure is a tree structure consisting of a series of processes in which the processor sequentially considers, compares, and calculates to diagnose the patient's mental illness,
The processor is
In each step of the decision-making structure, any one of the at least one second EEG characteristic is compared with a threshold value, and according to the comparison result, a next step of comparing the second EEG characteristic and the threshold value is determined,
a description and importance of the at least one second brain wave characteristic for understanding the decision structure is provided together with the decision structure, wherein the mental disorder the patient suffers from if the patient also suffers from a plurality of mental disorders provides information about the relative size,
The machine learning model uses the EEG by age and EEG characteristics for each channel included in the EEG by age as learning data for diagnosing mental disorders, uses a filter having a shared importance, and the decision-making structure visualized with the diagnosis result Use more feedback information from medical staff, including correction information for errors in
An explanatory artificial intelligence system in which the age of the patient, the mental state of the patient identified through the self-response questionnaire, and the health state information of the patient identified by the medical staff are additionally input to the machine learning model to perform mental disease diagnosis. delete delete delete delete delete delete In the method of providing information for diagnosing a mental illness performed by the explainable artificial intelligence system according to claim 1,
receiving the patient's brain waves;
pre-processing the received EEG through noise removal and epoching;
extracting at least one first EEG characteristic from the preprocessed EEG;
Using a machine learning model learned for diagnosing mental illness, at least one second EEG characteristic necessary for diagnosing the patient's mental illness among the at least one first EEG characteristic and the importance of each of the at least one second EEG characteristic determining;
generating a decision-making structure for diagnosing the patient's mental illness;
generating information for diagnosing the mental disorder of the patient by substituting the at least one second EEG characteristic and the importance into the decision-making structure; and
Visualizing and providing the decision-making structure as explanatory information for explaining the diagnosis result and the basis for the diagnosis;
The diagnosis result is information about at least one mental disorder suffering from the patient,
The description of the at least one second EEG characteristic includes a channel name required for diagnosing the patient's mental illness, an EEG type and EEG power in the channel, and a connection diagram between channels,
The decision-making structure is a tree structure consisting of a series of processes in which the processor sequentially considers, compares, and calculates to diagnose the patient's mental illness,
In each step of the decision-making structure, any one of the at least one second EEG characteristic is compared with a threshold value, and according to the comparison result, a next step of comparing the second EEG characteristic and the threshold value is determined,
providing a description and importance of the at least one second EEG characteristic for understanding the decision-making structure together with the decision-making structure;
When the patient suffers from a plurality of mental disorders, information on the mental illness the patient suffers from is provided as a relative value,
The machine learning model uses the EEG by age and EEG characteristics for each channel included in the EEG by age as learning data for diagnosing mental disorders, uses a filter having a shared importance, and the decision-making structure visualized with the diagnosis result using more feedback information from medical staff, including correction information for errors in
Information for diagnosing mental illness, in which the age of the patient, the mental state of the patient identified through the self-response questionnaire, and the health state information of the patient identified by the medical staff are additionally input to the machine learning model to perform the diagnosis of mental illness How to provide. delete A computer-readable recording medium storing a program for implementing the method of providing information for diagnosing a mental disorder according to claim 8.

Images