KR102239984B1 - Apparatus and method for providing information on epilepsy symptoms based on resting-state EEG - Google Patents

Abstract

The present invention relates to an apparatus and method for providing epilepsy diagnosis information for diagnosing whether or not there is epilepsy, and further, a degree of brain network change, using a patient's resting EEG, which is used to generate a plurality of resting EEG from pre-stored medical data. After obtaining, the brain wave data collection unit for classifying and storing according to the degree of change in the brain network; After measuring the patient's EEG, the EEG measurement unit for extracting and providing only the EEG corresponding to the resting period; A functional connectivity matrix generator for generating a functional connectivity matrix for each resting EEG stored in the EEG data collection unit when learning a neural network, and for generating a functional connectivity matrix for the patient EEG when analyzing a patient EEG; A graph theory analysis unit for grasping a graph theory analysis value by performing a graph theory analysis on the functional connectivity matrix; When learning a neural network, a graph theory analysis value corresponding to each resting EEG stored in the EEG data collection unit is taken as an input condition, and a plurality of learning data having a degree of brain network change as an output condition is generated, and then the plurality of learning. A neural network learning unit that repeatedly trains a neural network through data; And an epilepsy analyzer configured to grasp and report a degree of change in a brain network corresponding to a graph theory analysis value derived from the patient EEG using the neural network when analyzing the patient EEG.

Description

Apparatus and method for providing information on epilepsy symptoms based on resting-state EEG}

The present invention relates to an apparatus and method for providing epilepsy diagnosis information using quiescent EEG, and in particular, an apparatus for providing epilepsy diagnosis information capable of diagnosing whether or not there is epilepsy, and furthermore, the degree of brain network change using a patient’s quiescent EEG, and It's about the method.

The human brain is clearly a complex system and exhibits spatiotemporal dynamics. Among the techniques for investigating the human brain dynamics, the electroencephalogram (EEG) provides a temporal resolution in milliseconds and a direct measurement of cortical activity. Since EEG translocation was first recorded by Berger in 1929, EEG has been used in many clinical settings, such as brain epidemiology studies and schizophrenia diagnosis. In addition, EEG signals have been widely used clinically to investigate brain disorders. In the case of patients with epilepsy, monitoring of EEG is an important factor in confirming the condition of patients with epilepsy. Therefore, various methods using EEG have been proposed to detect and detect epilepsy.

Epilepsy is the second most common brain disease, and 1% of the world's population suffers from it. Epilepsy is a disease characterized by the occurrence of unpredictable and uncontrolled seizures. Many techniques have been studied to predict and control these epilepsy attacks in advance. One of these techniques is "On the predictability of epileptic seizures" published by F. Mormann et al., published in Clinical Neurophysiology: Official Journal of the International Federation of Clinical Neurophysiology in 2005.

According to this paper, the delta (0.5-4 Hz) band frequency component of the EEG signal measured in the brain of a patient with epilepsy in the prognostic period before the onset of epileptic seizures is relative to the frequency components of all other bands. It is explained that it decreases to.

However, this technique alone has the disadvantage of not being able to clearly detect the precursor section for all epileptic seizures. In addition, in the case of patients with low incidence of epilepsy seizures, there is also a disadvantage that diagnosis becomes very difficult even if they have epilepsy disease.

In order to solve the above problems, the present invention is to provide an apparatus and method for providing epilepsy diagnosis information to diagnose whether or not there is epilepsy, and furthermore, the degree of change in the brain network by using a patient's resting EEG. .

In addition, it is intended to provide an apparatus and method for providing epilepsy diagnosis information in which functional connectivity analysis is performed on the resting EEG of a normal person and a person with epilepsy, and deep-learning (or machine learning) the same for use in diagnosing epilepsy.

The object of the present invention is not limited to the above-mentioned object, and other objects not mentioned will be clearly understood by those of ordinary skill in the art from the following description.

As a means for solving the above problems, according to an embodiment of the present invention, after obtaining a plurality of resting EEG from pre-stored medical data, the EEG data collection unit for classifying and storing according to the degree of change in the brain network; After measuring the patient's EEG, the EEG measurement unit for extracting and providing only the EEG corresponding to the resting period; A functional connectivity matrix generator for generating a functional connectivity matrix for each resting EEG stored in the EEG data collection unit when learning a neural network, and for generating a functional connectivity matrix for the patient EEG when analyzing a patient EEG; A graph theory analysis unit for grasping a graph theory analysis value by performing graph theory analysis on the functional connectivity matrix; When learning a neural network, a graph theory analysis value corresponding to each resting EEG stored in the EEG data collection unit is taken as an input condition, and a plurality of learning data having a degree of brain network change as an output condition is generated, and then the plurality of learning A neural network learning unit that repeatedly trains a neural network through data; And an epilepsy analysis unit for grasping and notifying a degree of change in a brain network corresponding to a graph theory analysis value derived from the patient EEG using the neural network when analyzing a patient EEG. Can provide.

When the functional connectivity matrix generator is input to the resting EEG or the patient EEG, it filters the input EEG with delta, theta, alpha, beta, and gamma bands, and then the delta, theta, alpha, beta, and gamma band EEG. It is characterized by generating a functional connectivity matrix for each n×n (n is the number of channels of the electrode).

The graph theory analysis unit is characterized in that to obtain a graph theory analysis value for each EEG band by performing a graph theory analysis on an n×n functional connectivity matrix for each of the delta, theta, alpha, beta, and gamma EEG bands. .

The graph theory analysis value is characterized by including a clustering coefficient, a global efficiency, a characteristic path length, and a modularity.

As a means for solving the above problem, according to another embodiment of the present invention, after acquiring a plurality of dormant brain waves from pre-stored medical data, classifying and storing according to the degree of brain network change; Sequentially selecting the resting brain waves, generating a functional connectivity matrix for the selected resting brain waves, and then performing a graph theory analysis on the functional connectivity matrix to obtain a graph theory analysis value; Generating a plurality of training data having the graph theory analysis value as an input condition and a brain network change degree as an output condition, and then training a neural network through the plurality of training data; Extracting only the EEG corresponding to the resting period from the patient EEG when the patient's EEG is measured while the neural network is completely trained; Generating a functional connectivity matrix for the patient's brain waves and then performing the graph theory analysis to obtain a graph theory analysis value of the patient's brain waves; And generating and outputting diagnostic information having information on the determined degree of change in the brain network after determining the degree of change in the brain network corresponding to the graph theory analysis value of the patient EEG using the neural network. A method of providing epilepsy diagnosis information using may be provided.

The present invention makes it possible to diagnose whether or not there is epilepsy, and furthermore, the degree of brain network change using the patient's resting EEG.

In addition, by performing a functional connection analysis of the resting EEG of a normal person and a person with epilepsy disease, deep learning (or machine learning) and using it for epilepsy diagnosis, it is possible to perform epilepsy diagnosis more quickly and accurately.

1 is a diagram illustrating an apparatus for providing diagnosis information for epilepsy using resting brain waves according to an embodiment of the present invention.
2 is a diagram illustrating an example of an electrode arrangement method of an EEG measurement unit according to an embodiment of the present invention.
3 is a view for explaining a graph theory analysis value to be used in the apparatus for providing diagnosis information for epilepsy according to an embodiment of the present invention.
4 and 6 are diagrams for explaining a neural network learning method for diagnosing epilepsy according to an embodiment of the present invention.
7 is a diagram illustrating a method for diagnosing epilepsy according to an embodiment of the present invention.

Objects and effects of the present invention, and technical configurations for achieving them will become apparent with reference to the embodiments described below in detail together with the accompanying drawings. In describing the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the subject matter of the present invention, a detailed description thereof will be omitted.

In addition, terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to the intention or custom of users or operators.

However, the present invention is not limited to the embodiments disclosed below, and may be implemented in various different forms. The present embodiments are provided only to make the disclosure of the present invention complete, and to fully inform the scope of the invention to those skilled in the art to which the present invention pertains, and the present invention is defined by the scope of the claims. It just becomes. Therefore, the definition should be made based on the contents throughout the present specification.

1 is a diagram illustrating an apparatus for providing diagnosis information for epilepsy using resting brain waves according to an embodiment of the present invention.

Referring to FIG. 1, the apparatus for providing epilepsy diagnosis information of the present invention includes an EEG data collection unit 10, an EEG measurement unit 20, a functional connectivity matrix generation unit 30, a graph theory analysis unit 40, and a neural network learning. It may include a unit 50, and an epilepsy analysis unit 60, and the like.

The EEG data collection unit 10 connects to a medical data server (i.e., an external device) managed by a medical institution, a government institution, etc., and acquires only the resting EEG required for neural network learning from the medical data stored and managed by the medical data server. do. In addition, the acquired resting EEG is collected and stored according to the degree of change in the brain network.

For example, the degree of change in the brain network can be divided into early, middle, late, and normal, and it would be most preferable that the number of resting EEGs for each degree of brain network change are all similar.

The EEG measurement unit 20 measures the patient's EEG over a certain period of time, and then extracts and provides only EEG corresponding to the resting period. In other words, it is possible to perform both neural network learning and neural network analysis using resting EEG.

In addition, the EEG measurement unit 20 of the present invention may measure a patient EEG using all of the electrode arrangement methods according to known techniques. For example, as shown in Fig. 2, 19 electrodes (Fp1, Fp2, F7, F8, F3, F4, FZ, T3, T4, C3, C4, CZ, T5, T6, P3, P4, PZ, O1, O2) and two reference electrodes (FPZ, OZ) can be used to measure the patient's EEG, but this can be variously changed as needed.

In addition, by extracting and storing only the signal of the section in which spikes and artifacts are not included in the entire signal section of the EEG, it is possible to perform the subsequent EEG analysis operation.

The functional connectivity matrix generation unit 30 receives the resting EEG stored in the EEG data collection unit 10 when learning the neural network, and receives the patient EEG measured through the EEG measurement unit 20 when analyzing the patient EEG. Then, after extracting the EEG in the delta, theta, alpha, beta, and gamma bands from the input EEG, a functional connectivity matrix is generated and output for each of them.

The graph theory analysis unit 40 performs a graph theory analysis on each functional connectivity matrix to provide a clustering coefficient (CC), global efficiency (GE), average characteristic path length, and modularity. The analysis value of graph theory including (modularity) should be identified.

At this time, the clustering coefficient (CC) refers to the degree of local clustering or the complexity of the network, and this means that the number of observed connections between neighboring nodes (connected nodes) for one node as shown in Equation 1 is all possible. It can be calculated by dividing by the number of connections.

[Equation 1]

Where k _i is the ith node degree, N is the number of all nodes in the network, w _ij is the strength of connectivity between nodes i and j, w _ih is the strength of connectivity between nodes i and h, and w _jh is The strength of the connectivity between nodes j and h, t _i represents the geometric mean of each connectivity that forms a triangle including the i-th node. For reference, node degree is the sum of connectivity between one node and all other nodes.

Global efficiency (GE) refers to information propagation efficiency in the entire network, and may be calculated as an average of all node efficiencies as shown in Equation 2.

[Equation 2]

는 노드 i 와 j 간의 가장 짧은 가중치 경로, w_uv는 노드 u와 v간의 연결성 강도, a_uv는 A라는 인접 매트릭스(adjacent matrix)의 u행 v열을 의미한다. Here, E _glob is the overall efficiency, E _i is the node efficiency, d _ij is the shortest path length, f is a function that converts weight to length,

Is the shortest weight path between nodes i and j, w _uv is the strength of connectivity between nodes u and v, and a _uv is the u row v column of the adjacent matrix of A.

The average characteristic path length indicates the degree of overall routing efficiency of the network, and can be calculated by averaging the path lengths of all nodes.

Modularity refers to a hierarchical module structure of a functional connectivity matrix, and can be calculated according to Equation 3.

[Equation 3]

는 크로네커 델타(Kronecker delta)이다. Where l is the number of edges between nodes in the graph, A _ij is the adjacency matrix, k _i is the i-th node degree,

Is the Kronecker delta.

For reference, when performing a graph theory analysis on the resting EEG, various measures as shown in FIG. 3 are obtained, but in the present invention, only the above four variables are used for the diagnosis of epilepsy.

This is a result of performing graph theory analysis in each frequency band (i.e., delta, theta, alpha, beta, gamma band), and has a significance probability (p-value) of less than 0.05 in common in all frequency bands, and also epilepsy. This is because they are variables that reflect important brain network abnormalities in pathogenesis.

The neural network learning unit 50 is activated when learning a neural network, and first, a plurality of learning having a graph theory analysis value corresponding to each resting EEG stored in the EEG data collection unit as an input condition, and the degree of brain network change as an output condition. Create the data. In addition, by repeatedly learning the neural network through a large number of learning data, the correlation between the graph theory analysis value and the degree of network change of epilepsy patients is learned in the neural network.

The epilepsy analysis unit 60 is activated when analyzing the patient's EEG. This provides the graph theory analysis value derived from the patient's brain waves to the neural network, so that the neural network acquires and outputs the degree of change in the brain network corresponding to the currently input graph theory analysis value. In addition, diagnostic information including information on the degree of change in the brain network identified through a neural network is generated to provide audio-visual guidance to the user, or information for audio-visual guidance provided to an external device through a preset device wired or wireless communication network. To create and provide.

Hereinafter, a method of providing diagnosis information for epilepsy using resting EEG according to the present invention will be described in detail with reference to FIGS. 4 to 7.

4 is a diagram illustrating a neural network learning method for diagnosing epilepsy according to an embodiment of the present invention.

In step S110, when learning a neural network for diagnosing epilepsy is requested, the EEG data collection unit 10 accesses a preset medical data server in response thereto to obtain a plurality of dormant EEG signals, and then according to the degree of change in the brain network. Sort and save.

In step S120, one of a plurality of resting brain waves collected through step S110 is selected to generate learning data corresponding thereto. That is, while sequentially selecting a plurality of resting brain waves, a plurality of learning data corresponding thereto is generated.

In step S130, the functional connectivity matrix generation unit 30 uses the previously registered delta brain wave filter (0-4 Hz), theta brain wave filter (4-8 Hz), and alpha brain wave filter (8-Hz) for the resting brain wave selected through step S120. 12 Hz), beta EEG filter (12-20 Hz), and gamma EEG filter (20-40 Hz) to extract EEG in delta, theta, alpha, beta, and gamma bands.

In step S140, the functional connectivity matrix generation unit 30 generates a functional connectivity matrix n×n (n is the number of electrodes) for each of the brain waves in the delta, theta, alpha, beta, and gamma bands as shown in FIG. 5. In this case, the n×n functional connectivity matrix can be obtained by measuring and storing functional connectivity between different brain regions for each of the 19 brain regions. Functional connectivity between brain regions is a quantification of the coherence value of two EEGs measured in two different brain regions, and is in the range of 0 to 1, and when the synchronization is 1, the two EEGs are strongly synchronized. It means that the two brain waves are not correlated with each other, and that the synchronicity is close to zero.

In step S150, the graph theory analysis unit 40 performs a graph theory analysis on the n×n functional connectivity matrix for each of the delta, theta, alpha, beta, and gamma bands of EEG, so that the graph theory analysis value for each EEG band ( That is, the clustering coefficient (CC), global efficiency (GE), average characteristic path length, and modularity) are extracted.

In step S160, the neural network learning unit 50 obtains and stores learning data having a graph theory analysis value for each EEG band extracted through step S150 as an input condition and a degree of brain network change as an output condition.

For reference, it can be seen that the graph theory analysis values of the normal person and the epilepsy disease person have different occurrence patterns as shown in FIG. 6. Therefore, by generating learning data reflecting the correlation between the graph theory analysis value and the degree of brain network change (early, mid-term, late-stage, normal) based on the resting brain waves of the normal person and the epileptic disease patient of the present invention, and by learning the neural network, It enables the calculation of the degree of change in the brain network based on neural networks.

The above steps S120 to S160 are repeatedly performed until learning data for all of the plurality of resting brain waves are generated.

That is, in step S170, it is checked whether all of the plurality of resting brain waves are selected and learning data corresponding thereto is generated, and if not, the step S120 is re-entered.

On the other hand, if all of the plurality of resting EEG are selected, the process proceeds to step S180, and the neural network learning unit 50 iteratively learns the neural network through a plurality of learning data acquired so far, so that the graph theory analysis value and the brain network are applied to the neural network. Let the correlation of the degree of change be learned.

In this way, when the user requests the diagnosis of epilepsy in a state in which the neural network is completely trained, the epilepsy diagnosis operation is performed based on the neural network as shown in FIG. 7.

7 is a diagram illustrating a method for diagnosing epilepsy according to an embodiment of the present invention.

First, in step S210, after requesting the patient to take a comfortable posture in an environment where there is no sound and radio wave interference, close his mouth and eyes, the patient's EEG is measured through the EEG measurement unit 20. In addition, spikes and artifacts are not included in the patient's EEG, and only EEG corresponding to the resting period is extracted and output.

In step S220, the functional connectivity matrix generation unit 30 filters the patient's EEG through a delta EEG filter, a theta EEG filter, an alpha EEG filter, a beta EEG filter, and a gamma EEG filter. Only brain waves are extracted (S220).

In step S230, n×n (n is the number of electrode channels) functional connectivity matrix for each of the delta, theta, alpha, beta, and gamma band EEG bands is generated in the same manner as in step S122 described above (S230).

In step S240, graph theory analysis is performed on the n×n functional connectivity matrix for each of the delta, theta, alpha, beta, and gamma bands through the graph theory analysis unit 40 to obtain graph theory analysis values for each brain wave band. Let's extract it.

In step S250, the graph theory analysis value for each EEG band extracted through step S240 is input to the learned neural network, and the degree of network change corresponding to the patient's EEG through the neural network (i.e., early, mid-stage, late-stage, normal) or epilepsy. (That is, whether the degree of change in the brain network is greater than or equal to a preset degree) is determined and notified (S250).

The above description is merely illustrative of the technical idea of the present invention, and those of ordinary skill in the art to which the present invention pertains will be able to make various modifications and variations without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

Claims (5)

An EEG data collection unit for acquiring a plurality of resting EEG from pre-stored medical data and then classifying and storing the EEG according to the degree of change in the brain network;
After measuring the patient's EEG, the EEG measurement unit for extracting and providing only the EEG corresponding to the resting period;
A functional connectivity matrix generator for generating a functional connectivity matrix for each resting EEG stored in the EEG data collection unit when learning a neural network, and for generating a functional connectivity matrix for the patient EEG when analyzing a patient EEG;
A graph theory analysis unit for grasping a graph theory analysis value by performing a graph theory analysis on the functional connectivity matrix;
When learning a neural network, after generating a plurality of learning data having a graph theory analysis value corresponding to each resting EEG stored in the EEG data collection unit as an input condition, and the degree of brain network change as an output condition, the plurality of learning A neural network learning unit that repeatedly trains a neural network through data; And
When analyzing the patient's EEG, including an epilepsy analysis unit for grasping and notifying the degree of change in the brain network corresponding to the graph theory analysis value derived from the patient EEG using the neural network,
The graph theory analysis value includes modularity, which is a hierarchical module structure of the functional connectivity matrix,
The modularity is calculated by dividing the difference between the adjacent matrix and (the product of the i-th and j-th node strengths per number of connection lines between nodes) and the product of the Kronecker delta by the number of connection lines between nodes in the graph. An apparatus for providing epilepsy diagnosis information using resting brain waves, characterized in that. The method of claim 1, wherein the functional connectivity matrix generation unit
When the resting EEG or the patient EEG is input, the input EEG is filtered with EEG in the delta, theta, alpha, beta, and gamma bands, and then n×n for each of the delta, theta, alpha, beta, and gamma bands. (n is the number of channels of the electrode) A device for providing epilepsy diagnosis information using resting brain waves, characterized in that it generates a functional connectivity matrix. The method of claim 2, wherein the graph theory analysis unit
EEG using a resting EEG, characterized in that by performing a graph theory analysis on the n×n functional connectivity matrix for each of the delta, theta, alpha, beta, and gamma bands of EEG, obtaining a graph theory analysis value for each EEG band. Device for providing diagnosis information. The method of claim 1 or 3, wherein the graph theory analysis value is
An apparatus for providing epilepsy diagnosis information using a resting EEG, further comprising a clustering coefficient, a global efficiency, and a characteristic path length. Acquiring a plurality of resting EEG from the pre-stored medical data by the EEG data collection unit, and then classifying and storing them according to the degree of change in the brain network;
Generating a functional connectivity matrix for the selected resting EEG, while a functional connectivity matrix generator sequentially selecting the resting EEG, and then performing a graph theory analysis on the functional connectivity matrix to obtain a graph theory analysis value;
Generating a plurality of training data having the graph theory analysis value as an input condition and a brain network change degree as an output condition, and then training a neural network through the plurality of training data;
When the patient's EEG is measured while the neural network is trained, the EEG measuring unit extracts only the EEG corresponding to the resting period from the patient EEG;
Obtaining a graph theory analysis value of the patient EEG by the functional connectivity matrix generation unit generating a functional connectivity matrix for the patient EEG, and then performing the graph theory analysis by a graph theory analysis unit; And
Including the step of generating and outputting diagnostic information having information on the determined degree of change in the brain network after the epilepsy analysis unit determines the degree of change in the brain network corresponding to the graph theory analysis value of the patient's EEG by using the neural network. and,
The graph theory analysis value includes modularity, which is a hierarchical module structure of the functional connectivity matrix,
The modularity is calculated by dividing the difference between the adjacent matrix and (the product of the i-th and j-th node strengths per number of connection lines between nodes) and the product of the Kronecker delta by the number of connection lines between nodes in the graph. A method of providing diagnosis information for epilepsy using resting brain waves, characterized in that.

Images