Information processing method, information processing program, and information processing device.

Abstract

This provides novel methods for analyzing measured brainwaves. [Solution] The information processing method comprises the steps of: acquiring electroencephalograms measured from a subject; generating state transitions between microstates by pattern matching the spatial patterns shown by the acquired electroencephalograms at predetermined intervals with a predetermined number of microstate templates; and reducing the state transitions between microstates to a predetermined number of transition components by clustering them, wherein each of the predetermined number of transition components includes some of the transition pairs from a plurality of types of transition pairs defined as state transitions.

Description

Technical Field

[0001] The present invention relates to an information processing method, an information processing program, and an information processing device.

Background Art

[0002] As a method for non-invasively measuring brain activity, typically, a method for measuring an electroencephalogram (hereinafter also abbreviated as "EEG") indicating the electrical activity of the brain is known. A method for evaluating the state of the brain activity of a subject based on the EEG measured from the subject has been proposed.

[0003] For example, a method for calculating a biomarker for a specific disease based on the time waveform or power spectrum of EEG is known.

[0004] A method for calculating a biomarker using EEG microstates, which are the minimum units indicating the state of brain activity, is also known. For example, a method for calculating biomarkers for mood disorders and anxiety disorders by meta-analysis of EEG microstates is known (for example, Non-Patent Document 1). Also, a method for calculating biomarkers based on the duration, occurrence frequency, state transition probability, etc. of EEG microstates is known (for example, Non-Patent Document 2).

Prior Art Documents

Non-Patent Documents

[0005]

Non-Patent Document 1

Non-Patent Document 2

[0006] In the background technologies described above, there are no biomarkers that can be used universally for a variety of diseases. Furthermore, simply evaluating the duration and state transition probability of EEG microstates is insufficient to calculate biomarkers that adequately reflect the state of brain activity.

[0007] This invention provides a novel method for analyzing measured electroencephalograms (EEGs). [Means for solving the problem]

[0008] An information processing method according to one embodiment of the present invention includes the steps of: acquiring electroencephalograms measured from a subject; generating state transitions between microstates by pattern matching the spatial patterns shown by the acquired electroencephalograms at predetermined intervals with a plurality of predetermined microstate templates; and reducing the state transitions between microstates to a predetermined number of transition components by clustering them. Each of the predetermined number of transition components includes a subset of transition pairs from a plurality of types of transition pairs defined as state transitions.

[0009] The contraction step may include determining a predetermined number of principal component axes included in the set of transition pairs included in the state transitions between microstates, and determining a cluster by collecting one or more transition pairs whose contribution to each principal component axis satisfies predetermined conditions for each of the determined predetermined number of principal component axes.

[0010] The predetermined conditions may include at least one of the following: the degree of contribution to the principal component axis of interest is greater than the contribution to the other principal component axes; and the degree of contribution to the principal component axis of interest is greater than a predetermined value.

[0011] The contraction step may include a step of clustering the set of transition pairs included in the state transitions between microstates using the k-means method or hierarchical clustering.

[0012] Multiple microstate templates may include a pair of templates with reversed polarity. State transitions between microstates may be based on microstates that take polarity into account.

[0013] The information processing method may further include the step of calculating a biomarker based on at least some of a predetermined number of transition components.

[0014] The steps for calculating a biomarker may include: acquiring a second electroencephalogram (EEG) measured from a subject for whom the biomarker is to be calculated; generating state transitions between the second microstates by pattern matching the spatial pattern shown by the second EEG at predetermined intervals with a predetermined number of microstate templates; and calculating the degree of involvement of at least some of the transition components among a predetermined number of transition components by clustering the state transitions between the second microstates.

[0015] Depending on the indicator items shown by the biomarker, at least some of the transition components may be determined.

[0016] According to another embodiment of the present invention, an information processing program is provided for causing a computer to execute the above-described information processing method.

[0017] An information processing apparatus according to another embodiment of the present invention includes means for generating a state transition between microstates by pattern matching between a spatial pattern for each predetermined period indicated by an electroencephalogram measured from a subject and a plurality of predetermined microstate templates, and means for reducing the state transitions between microstates to a predetermined number of transition components by clustering. Each of the predetermined number of transition components includes some of the transition pairs of a plurality of types of transition pairs defined as state transitions.

Advantages of the Invention

[0018] According to the present embodiment, a novel analysis method for an electroencephalogram to be measured can be realized.

Brief Description of the Drawings

[0019] [Figure 1] It is a schematic diagram showing an overall configuration example including an information processing apparatus according to the present embodiment. [Figure 2] It is a schematic diagram showing a hardware configuration example of the information processing apparatus according to the present embodiment. [Figure 3] It is a flowchart showing an example of a processing procedure of an information processing method according to the present embodiment. [Figure 4] It is a diagram for explaining a calculation process of a vector in the information processing method according to the present embodiment. [Figure 5] It is a schematic diagram showing an example of a spatial pattern in the information processing method according to the present embodiment. [Figure 6] It is a schematic diagram showing an example of a state space in which microstates are associated in the information processing method according to the present embodiment. [Figure 7] It is a diagram for explaining an example of clustering for extracting a transition component in the information processing method according to the present embodiment. [Figure 8] It is a diagram showing an example of a transition component extracted by the clustering shown in FIG. 7. [Figure 9] It is a diagram for explaining another example of clustering for extracting a transition component in the information processing method according to the present embodiment. [Figure 10] This is a diagram for explaining another example of clustering for extracting transition components in the information processing method according to this embodiment. [Figure 11] This is a diagram showing an example of evaluating the performance as a biomarker of aging. [Figure 12] This is a diagram showing an example of evaluating the performance as a biomarker of schizophrenia. [Figure 13] This is a schematic diagram showing an example of a biomarker definition used in the information processing method according to this embodiment. [Embodiment for Carrying Out the Invention]

[0020] Embodiments of the present invention will be described in detail with reference to the drawings. For the same or corresponding parts in the drawings, the same reference numerals are given and the description thereof will not be repeated.

[0021] [A. Overall Configuration Example] First, an overall configuration example including an information processing apparatus according to this embodiment will be described.

[0022] FIG. 1 is a schematic diagram showing an overall configuration example including an information processing apparatus 100 according to this embodiment. Referring to FIG. 1, the information processing apparatus 100 acquires electroencephalogram measured from a subject 2. Hereinafter, an example using EEG data as an example of electroencephalogram will be described. [[ID=3​​​​​The multiplexer 42 sequentially selects multiple channels (electrodes 12) output from the cap 10 and electrically connects them to the noise filter 44. The noise filter 44 is a filter that allows only specific frequency components to pass through, removing noise components contained in the signal (electrical signal) appearing in the selected channel. For example, the passband can be set to 0.016 to 250 Hz.

[0025] The A / D converter 46 samples the electrical signal (analog signal) output from the noise filter 44 at predetermined intervals and outputs it as a digital signal.

[0026] Figure 1 shows an example configuration in which the measurement circuit 40 and the information processing device 100 are directly connected. However, a recording device for collecting EEG data may be connected to the measurement circuit 40, and the information processing device 100 may acquire the EEG data collected by the recording device by any method (for example, via a recording medium).

[0027] Figure 2 is a schematic diagram showing an example of the hardware configuration of the information processing device 100 according to this embodiment. The information processing device 100 may be configured using, for example, a personal computer following a general-purpose architecture.

[0028] The information processing device 100 includes, as its main components, a processor 102, a memory 104, an input unit 106, a network controller 108, a measurement interface 110, a display unit 112, and a storage unit 120.

[0029] The processor 102 consists of arithmetic processing circuits such as a CPU (Central Processing Unit) and a GPU (Graphical Processing Unit), and executes the code contained in various programs stored in the storage 120 in a specified order to realize the various functions described later. The memory 104 consists of DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory), and holds the code of the programs executed by the processor 102 and various work data necessary for program execution.

[0030] The input unit 106 typically consists of a mouse or keyboard, and accepts user input.

[0031] The network controller 108 exchanges data with external devices (for example, a data server on the cloud). The network controller 108 is composed of any communication components such as wired LAN (Local Area Network), wireless LAN, USB (Universal Serial Bus), and Bluetooth (registered trademark).

[0032] The measurement interface 110 receives the EEG (digital signal) output from the measurement circuit 40.

[0033] The display unit 112 displays the processing results from the processor 102, etc. The display unit 112 consists of a liquid crystal display or the like.

[0034] The storage 120 consists of, for example, a hard disk or an SSD (Solid State Drive) and holds various programs executed by the processor 102, various data necessary for processing, and setting values. More specifically, the storage 120 includes the OS (Operating System) 122, an information processing program 124, EEG data 126, a microstate template 128, and a biomarker definition 130.

[0035] The information processing program 124 includes instructions for performing analysis processing on the EEG data 126, as described later.

[0036] The microstate template 128 is a map for determining EEG microstates. The microstate template 128 may also be part of a trained model.

[0037] Biomarker definition 130 includes a correspondence that shows one or more significant transition components for each indicator.

[0038] [B. Processing Procedure] One objective of the information processing method according to this embodiment is to provide an index that represents the spatiotemporal dynamics of electroencephalograms (EEGs). In other words, the information processing method according to this embodiment quantifies the transitions in EEG state dynamics. The quantified transitions in state dynamics can be used as biomarkers for various diseases.

[0039] Figure 3 is a flowchart showing an example of the processing procedure for an information processing method according to this embodiment. Each step shown in Figure 3 may be realized by the processor 102 of the information processing device 100 executing the information processing program 124.

[0040] Referring to Figure 3, the information processing device 100 acquires EEG data measured from one or more subjects (step S2). The acquired EEG data includes time waveforms showing signal intensity for each predetermined number of channels (e.g., 32 channels). The length of the time-axis direction of the time waveforms does not have to be the same among the EEG data.

[0041] The information processing device 100 selects one of the acquired EEG data (step S4). The information processing device 100 extracts an interval of a predetermined length as an epoch from the selected EEG data (step S6). Multiple epochs may be extracted from a single EEG data.

[0042] The information processing device 100 selects one epoch (step S8). For the selected epoch, the information processing device 100 calculates a vector in channel space (having the dimension of the number of channels) consisting of the signal intensity of each channel at predetermined periods on the time axis (hereinafter also referred to as the "sampling period") (step S10), and calculates time-series data of the vector for one epoch (step S12).

[0043] Furthermore, if the sampling period used to calculate the vector is longer than the collection period of the EEG data (raw data), multiple signal intensities may exist for each channel within a single sampling period used to calculate the vector. In such cases, a representative value may be calculated based on the multiple signal intensities included in the sampling period used to calculate the vector, and the vector may be calculated based on this calculated representative value.

[0044] Figure 4 is a diagram illustrating the vector calculation process in the information processing method according to this embodiment. Figure 4 shows an example of EEG data for one epoch (1000 ms). A vector is calculated for each sampling period (the area enclosed in the frame in Figure 4). The sampling period may be, for example, 4 ms.

[0045] Figure 4 shows both the electroencephalography, which represents the distribution of brain potentials corresponding to the electrode positions, and the determined spatial pattern.

[0046] Referring again to Figure 3, the information processing device 100 determines whether or not time-series data for the vector has been calculated for all epochs (step S14). If there are any epochs for which time-series data for the vector has not been calculated (NO in step S14), the process from step S8 onwards is repeated.

[0047] If time-series vector data has been calculated for all epochs (YES in step S14), the information processing device 100 determines whether all EEG data has been processed (step S16). If there is any unprocessed EEG data remaining (NO in step S16), the process from step S4 onwards is repeated.

[0048] If all EEG data has been processed (YES in step S16), the information processing device 100 determines the number of EEG microstates (hereinafter also simply referred to as "microstates") and the range of each microstate in channel space based on the time-series data of the calculated vectors for each epoch (step S18).

[0049] The processing in steps S2 to S18 corresponds to the process of determining a microstate template for determining the microstates in each sampling period (a type of learning process).

[0050] Next, the information processing device 100 determines the microstates at each sampling period of each acquired EEG data using the determined microstate template (step S20). Specifically, the information processing device 100 generates state transitions between microstates by pattern matching the spatial pattern shown by the EEG data for each sampling period with a predetermined set of microstate templates.

[0051] The processing in step S20 generates time-series data of microstates corresponding to the EEG data. By determining the microstates at each sampling period, discretized state transitions between microstate templates are generated. These discretized state transitions represent continuous EEG state transitions in space and time.

[0052] The information processing device 100 clusters the calculated discretized state transitions (step S22) and extracts the transition components (step S24). At least steps S22 and S24 correspond to the extraction of transition components by reduction. That is, the information processing device 100 reduces the state transitions between microstates to a predetermined number of transition components by clustering them.

[0053] The information processing device 100 acquires EEG data measured from the subject for which the index is to be calculated (step S30). The information processing device 100 determines the microstates at each sampling period of the acquired EEG data using the determined microstate template (step S32). The information processing device 100 clusters the microstates at each determined sampling period, i.e., the discretized state transitions (step S34). Then, based on the clustering results, the information processing device 100 calculates the degree of involvement for at least some of the transition components (step S36). The calculated degree of involvement may be used as an index for the subject in question. The degree of involvement will be described later.

[0054] In this way, the information processing device 100 calculates a biomarker, which is an example of an index, based on at least some of the transition components among a predetermined number of transition components. The calculated index is based on the spatiotemporal dynamics of electroencephalograms. The calculated index may be output in any way. For example, the calculated index may be displayed on the display unit 112 of the information processing device 100, or stored in the storage 120 of the information processing device 100. The calculated index may also be transmitted to other information processing devices or servers.

[0055] Note that the processes in steps S2 to S24 and steps S30 to S36 do not need to be executed simultaneously. Only the processes in steps S30 to S36 may be repeatedly executed using the microstate template and extracted transition components determined by the processes in steps S2 to S24.

[0056] [C. Determining Microstate Templates] Next, we will explain the determination of the microstate template shown in Figure 3 (steps S2 to S18 in Figure 3).

[0057] Microstates can be considered as a range of stable states (attractors). By assuming that brain states transition between microstates, it becomes possible to observe the transitions between brain states labeled by microstates.

[0058] The information processing device 100 determines multiple clusters by clustering numerous vectors generated from each epoch of EEG data. Each determined cluster corresponds to a microstate that represents a unique brain state. Known methods such as modified k-means clustering and hierarchical clustering can be used for clustering.

[0059] Vectors belonging to the same cluster have similar characteristics. Therefore, the spatial patterns generated from vectors belonging to the same cluster will also be similar. The spatial pattern maps the signal intensity of each channel represented by each vector according to the location of the electrode corresponding to each channel (i.e., the brain region where the information is collected). The spatial pattern corresponding to each cluster (i.e., microstate) can also be called an eigenspace pattern.

[0060] Figure 5 is a schematic diagram showing an example of a spatial pattern in the information processing method according to this embodiment. Figure 5(A) shows examples of five microstates A, B, C, D, and E as brain states. A microstate template is generated corresponding to each microstate. The microstate template is used for pattern matching with the spatial pattern shown by the vector calculated at each sampling period.

[0061] In the information processing method according to this embodiment, each determined microstate template (original template) and an inverse polarity template obtained by reversing the polarity of the original template are used.

[0062] Here, we define a microstate with a positive overall potential on the anterior (face) side as "Positive," and a microstate with a negative overall potential on the anterior (face) side as "Negative."

[0063] Figure 5(B) shows 10 microstates generated from the 5 microstates shown in Figure 5(A) (5 positive and 5 negative). For example, in the modified k-means method, polarity is ignored, so microstates are clustered as shown in Figure 5(A), but by reversing the polarity of each, the brain state can be estimated more accurately. Thus, multiple microstate templates include a pair of templates with reversed polarity.

[0064] In the following explanation, of the five microstates A, B, C, D, and E, positive states will be indicated with a "+" (e.g., A+) and negative states with a "-" (e.g., A-).

[0065] Based on microstates that take polarity into account (microstate templates), continuous spatiotemporal EEG state transitions are calculated from the EEG data measured from subject 2. In other words, the state transitions between microstates are based on microstates that take polarity into account.

[0066] To represent continuous EEG state transitions (continuous transitions in state dynamics) in space and time, a state space may be introduced. By introducing a state space, continuous EEG state transitions can be represented in space and time. Such a state space is sometimes called a neural manifold.

[0067] Figure 6 is a schematic diagram showing an example of a state space associated with microstates in an information processing method according to this embodiment. Referring to Figure 6, each microstate corresponds to a predetermined region in the state space.

[0068] [D. Calculation of EEG state transitions and extraction of transition components] Next, we will explain the calculation of electroencephalogram (EEG) state transitions and the extraction of transition components through reduction (steps S20-S24 in Figure 3).

[0069] A continuous EEG state transition in space and time can be discretized into state transitions between microstates (see Figure 6). In other words, a continuous trajectory is discretized as movement between regions. As described above, if we define 10 microstates, there are 10 × 10 possible (discrete) transitions from one microstate in one sampling period to another in the next sampling period.

[0070] The information processing device 100 uses a microstate template to determine the microstates at each sampling period in each epoch of the EEG data. Based on this, the information processing device 100 calculates time-series data of the microstates in one epoch (e.g., A+, C+, E-, ...). A correlation value is calculated between the spatial pattern derived from the vectors at each sampling period and each microstate template. The microstate at each sampling period is determined based on the microstate template showing the highest correlation value. Time-series data of the microstates is then calculated from the microstates determined for each sampling period.

[0071] The information processing device 100 determines multiple clusters by clustering discrete transitions contained in time-series data of microstates calculated based on EEG data measured from one or more subjects, based on covariance relationships.

[0072] For clustering, known methods such as principal component analysis with varimax rotation (a method that determines item groups according to their contribution to the principal component axes), k-means method, and hierarchical clustering (Ward method) can be used.

[0073] Figure 7 is a diagram illustrating an example of clustering for extracting transition components in an information processing method according to this embodiment. Figure 8 is a diagram showing an example of transition components extracted by the clustering shown in Figure 7.

[0074] In Figures 7 and 8, principal component analysis with varimax rotation is used to reduce the data to eight principal components, and as an example, clusters are determined by collecting transition pairs that show the maximum contribution to each principal component axis. More specifically, in this principal component analysis, a process is performed to determine a predetermined number of principal component axes included in the set of transition pairs included in the state transitions between microstates, and for each of the determined predetermined number of principal component axes, a process is performed to collect one or more transition pairs whose contribution to each principal component axis satisfies predetermined conditions and determine clusters.

[0075] Figure 7(A) shows a graph in which the components are arranged in order from those with the highest eigenvalues. As shown in Figure 7(A), eight principal components (PCs) with significant eigenvalues ​​are extracted. As shown in Figure 7(B), the extracted eight principal components account for 87.65% of the variance of the entire original data.

[0076] Figure 8 shows eight clustering examples, where transition pairs showing the maximum contribution to each extracted principal component axis are grouped together. In Figure 8, uppercase letters (A, B, C, D, E) indicate positive, and lowercase letters (a, b, c, d, e) indicate negative.

[0077] Figure 9 illustrates another example of clustering for extracting transition components in the information processing method according to this embodiment. Figure 9 shows an example of processing in which transition components are extracted using the k-means method (fixed at k=8). The maximum number of iterations was set to 1000, and the minimum number of iterations to 50.

[0078] Figure 9(A) shows an example of the results of silhouette analysis using the k-means method. Figure 9(B) shows the transition components extracted by silhouette analysis. Among the transition components shown in Figure 9, those that are the same as the extracted transition components shown in Figure 8 are shown in bold.

[0079] Figure 10 illustrates yet another example of clustering for extracting transition components in the information processing method according to this embodiment. Figure 10 shows an example of processing in which transition components have been extracted by hierarchical clustering.

[0080] Figure 10(A) shows an example of the results of hierarchical clustering. Figure 10(B) shows the transition components extracted by hierarchical clustering. Among the transition components shown in Figure 10, those that are the same as the extracted transition components shown in Figure 8 are shown in bold.

[0081] As shown in Figures 9 and 10, the process of reducing to a predetermined number of transition components may include a process of clustering the set of transition pairs included in the state transitions between microstates using the k-means method or hierarchical clustering.

[0082] In this way, a predetermined number of transition components (for example, 8) are extracted from 10 x 10 possible transitions. Similar transition components are extracted regardless of the clustering method used. These extracted transition components can be considered unique transition components.

[0083] Here, we will explain in more detail an example of the process for extracting transition components. Each of the predetermined number of transition components contains some of the transition pairs (10 × 10 in the example above) that are defined as state transitions between microstates. A single transition pair may belong to only one transition component, or it may belong to multiple transition components. Furthermore, each of the predetermined number of transition components may contain one or more transition pairs selected from the multiple types of transition pairs defined as state transitions, according to the loading of each transition pair for each transition component. Note that there may also be transition pairs that do not belong to any transition component.

[0084] For each transition pair, it may be determined which transition component it belongs to based on the degree of its contribution to each principal component axis. In other words, in the process of extracting transition components by reduction, for each of the predetermined number of principal component axes, one or more transition pairs whose contribution to each principal component axis satisfies predetermined conditions are collected to determine a cluster (transition component). As described above, any clustering method may be used.

[0085] The predetermined condition may be that the degree of contribution to the principal component axis of interest is greater than the degree of contribution to other principal component axes. For example, each transition pair may belong to the single principal component axis in which the degree of contribution of that transition pair is maximized. Alternatively, each transition pair may belong to multiple principal component axes corresponding to the higher degrees of contribution shown by that transition pair.

[0086] Alternatively, the predetermined condition may be that the degree of contribution to the principal component axis of interest is greater than a predetermined value. For example, under this condition, each transition pair belongs to one or more principal component axes in which the degree of contribution of the transition pair exceeds a predetermined value (threshold).

[0087] Next, we will describe an example of an extracted transition component. "Hub C" consists of transition pairs between microstate C (C+ and C-) and other microstates of opposite polarity A, B, D, E (A-, B-, D-, E- and A+, B+, D+, E+), as well as self-transitions of microstate C (C+ and C-). In other words, Hub C contains transition pairs centered around microstate C.

[0088] In "Hub E," most of the contributing transition pairs are transition pairs that originate from or end with microstates E (E+ and E-).

[0089] In the "ECD rotation," most of the contributing transition pairs belong to a route that transitions in a specific direction in the order of microstates E+, C+, D+, E-, C-, D-.

[0090] The "AD plane" consists of transition pairs between microstates A (A+ and A-) and microstates D (D+ and D-).

[0091] The "BD plane" consists of transition pairs between microstates B (B+ and B-) and microstates D (D+ and D-).

[0092] The "DCE rotation" consists of transition pairs that belong to a route that transitions in a specific direction in the order of microstates D+, C+, E+, D-, C-, E- (i.e., the reverse direction of the ECD rotation).

[0093] In the "ACB rotation," most of the contributing transition pairs belong to a route that transitions in a specific direction in the order of microstates A+, C+, B+, A-, C-, B-.

[0094] The "BCA rotation" consists of transition pairs that belong to a route (i.e., the reverse direction of the ACB rotation) in which the microphone states B+, C+, A+, B-, C-, A- transition in a specific direction.

[0095] Each of these contracted transition components is a set of transition pairs that share unique characteristics. [E. Calculation of indicators] Next, we will explain the process for calculating indicators (steps S30 to S36 in Figure 3). As an example of an indicator, we will explain an example of the process for calculating biomarkers that indicate the subject's condition or the possibility of disease.

[0096] The information processing device 100 calculates time-series data of microstates from EEG data measured from subjects for which indicators are to be calculated. Specifically, the information processing device 100 acquires EEG data (electroencephalogram) measured from subjects for which biomarkers are to be calculated, and generates state transitions between microstates by pattern matching the spatial patterns shown by the acquired EEG data at predetermined intervals with a plurality of predetermined microstate templates.

[0097] The information processing device 100 calculates the degree of involvement for at least some of a predetermined number of reduced transition components based on time-series data of state transitions between microstates. The degree of involvement may be an index indicating the extent to which each transition pair included in the state transitions shown by the EEG data measured from the subject is involved in each transition component. Alternatively, the degree of involvement may be an index indicating the extent to which each transition component is significantly present in the state transitions shown by the EEG data measured from the subject.

[0098] For example, the degree of involvement with a transition component of interest may be calculated based on the sum of the component scores of one or more transition pairs belonging to that transition component, or based on the sum of the contribution rates of one or more transition pairs belonging to that transition component. Alternatively, the degree of involvement with a transition component of interest may be calculated using both component scores and contribution rates. Furthermore, the method for calculating the degree of involvement may be any method.

[0099] In this way, the information processing device 100 calculates biomarkers, which are indicators that show the characteristics of a subject or the likelihood of disease, based on the eigenvalues ​​of one or more transition components.

[0100] The following explains how it can be used as a biomarker. Figure 11 shows an example of evaluating the performance of aging as a biomarker. Figure 11 shows an example of the results of a multivariate logistic regression analysis that examined whether eight transition components predict two age groups, one belonging to an older group and the other to a younger group. For transition components where most transition pairs have negative weight values, the signs of the varimax-rotated transition component scores were reversed to align the relationships between transition components of the age groups (hub E, BD plane, ACB rotation, BCA rotation).

[0101] As shown in Figure 11, all eight transition components exhibit sufficient characteristics to distinguish between the older and younger groups. In particular, the transition components of hub E, AD plane, BD plane, DCE rotation, and ACB rotation show relatively strong characteristics for group identification.

[0102] Figure 12 shows an example of evaluating the performance of a biomarker for schizophrenia. Similar to Figure 11, Figure 12 shows an example of the results of a multivariate logistic regression analysis that examined whether the eight transition components predict the disease group between subjects showing a tendency toward schizophrenia and other subjects.

[0103] As shown in Figure 12, all eight transition components exhibit sufficient features to distinguish between the two groups. In particular, the transition components of the AD plane, ACB rotation, and BCA rotation exhibit relatively strong features for distinguishing between the groups.

[0104] Thus, depending on the indicator items shown by the biomarker, at least some of the transition components used to calculate the biomarker may be determined from a predetermined number of transition components.

[0105] As shown in Figures 12 and 13, a combination of involvement levels (one or more dimensional values) for one or more transition components may be used as a biomarker. Alternatively, a value (one-dimensional) obtained by multiplying the involvement levels for one or more transition components by corresponding weighting coefficients and adding them together may be used as a biomarker. The respective weighting coefficients may be determined based on factors such as the degree to which the corresponding transition component is significant as a biomarker.

[0106] [F. Operation example] Next, an example of the operation of the information processing method according to this embodiment will be described.

[0107] The microstate template determined by the process described above can be used in common among subjects. Therefore, the determined microstate template (microstate template 128 shown in Figure 2) may be provided to other information processing devices.

[0108] In addition, a biomarker definition 130 that defines one or more significant transition components for each indicator item from among the abbreviated transition components may be provided to other information processing devices.

[0109] Figure 13 is a schematic diagram showing an example of a biomarker definition 130 used in the information processing method according to this embodiment. Referring to Figure 13, the biomarker definition 130 has one or more transition components (eight in this example) that are used to calculate the biomarker, associated with each of the one or more indicator items.

[0110] For example, for indicator item 1, it is defined that the biomarker is calculated using the transition components of hub E, AD plane, BD plane, DCE rotation, and ACB rotation.

[0111] In addition, while the biomarker definition 130 in Figure 13 specifies the transition components (i.e., effective / ineffective) used in calculating the biomarker, a weighting coefficient may be set for each transition component. For transition components that are not significant in calculating the biomarker, zero or a value close to zero may be set as the weighting coefficient.

[0112] By performing evaluations for each indicator item as shown in Figures 11 and 12, biomarker definitions can be generated.

[0113] The microstate templates and biomarker definitions may be made downloadable from the server, or they may be made available to other information processing devices in any way.

[0114] In this way, by using microstate templates and biomarker definitions, biomarkers can be calculated based on EEG data measured from subjects.

[0115] [G. Others] In the above description, an example is shown in which a biomarker is calculated using the measured EEG data essentially as is. However, a biomarker may also be calculated by extracting transition components of a specific frequency band included in the EEG data, or transition components for each frequency band. For example, frequency bands such as 1-4Hz (δ band), 4-8Hz (θ band), 8-12Hz (α band), 13-30Hz (β band), and 30-45Hz (γ band) are known for EEG data, and transition components may be extracted for some or all of these frequency bands. Therefore, in the information processing method according to this embodiment, state transitions may be defined based on broadband (e.g., 2-20Hz) EEG data, or state transitions may be defined by focusing on an arbitrary frequency band. When defining state transitions for each of multiple frequency bands, the transition components extracted for each frequency band may be combined and used as a biomarker.

[0116] In the explanation above, an example was given using EEG, but the method is also applicable when using magnetoencephalography (MEG).

[0117] [H. Advantages] According to this embodiment, while maintaining information on the spatiotemporal dynamics of electroencephalograms (EEGs), biomarkers, which are biological indicators of a subject, can be calculated by focusing on multiple characteristic transition components included in said spatiotemporal dynamics. The biomarkers calculated in this way comprehensively consider both spatial and temporal brain activity.

[0118] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is indicated by the claims rather than by the description of the embodiments above, and all modifications within the meaning and scope equivalent to the claims are intended to be included. [Explanation of Symbols]

[0119] 2 subjects, 10 caps, 12 electrodes, 40 measurement circuits, 42 multiplexers, 44 noise filters, 46 A / D converters, 100 information processing units, 102 processors, 104 memory, 106 input units, 108 network controllers, 110 measurement interfaces, 112 display units, 120 storage, 124 information processing programs, 126 EEG data, 128 microstate templates, 130 biomarker definitions.

Claims

1. The steps include obtaining brainwaves measured from the subject, The steps include generating state transitions between microstates by pattern matching between the spatial pattern shown by the acquired electroencephalogram at predetermined intervals and a predetermined number of microstate templates, An information processing method comprising the step of reducing the state transitions between the microstates to a predetermined number of transition components by clustering them, wherein each of the predetermined number of transition components includes a portion of the transition pairs of a plurality of types defined as state transitions.

2. The aforementioned step of contraction is, The steps include determining a predetermined number of principal component axes included in the set of transition pairs included in the state transitions between the microstates, The information processing method according to claim 1, comprising the step of determining a cluster by collecting one or more transition pairs whose contribution to each principal component axis satisfies predetermined conditions for each of the predetermined number of principal component axes determined.

3. The information processing method according to claim 2, wherein the predetermined conditions include at least one of the following: the degree of contribution to the principal component axis of interest is greater than the contribution to other principal component axes, and the degree of contribution to the principal component axis of interest is greater than a predetermined value.

4. The information processing method according to claim 1, wherein the contraction step includes a step of clustering the set of transition pairs included in the state transitions between the microstates using the k-means method or hierarchical clustering.

5. The plurality of microstate templates include a pair of templates with reversed polarity, The information processing method according to any one of claims 1 to 3, wherein the state transitions between the microstates are based on microstates that take polarity into consideration.

6. The information processing method according to any one of claims 1 to 3, further comprising the step of calculating a biomarker based on at least some of the predetermined number of transition components.

7. The step of calculating the biomarker is: The steps include: obtaining a second electroencephalogram measured from a subject for whom the biomarker is to be calculated; The process involves generating state transitions between second microstates by pattern matching between the spatial pattern shown by the second electroencephalogram at predetermined intervals and a predetermined set of microstate templates, and The information processing method according to claim 6, comprising the step of calculating the degree of involvement of at least some of the predetermined number of transition components by clustering the state transitions between the second microstates.

8. The information processing method according to claim 7, wherein at least some of the transition components are determined depending on the indicator items indicated by the biomarker.

9. An information processing program for causing a computer to execute the information processing method described in any one of claims 1 to 3.

10. A means for generating state transitions between microstates by pattern matching between spatial patterns shown at predetermined intervals by electroencephalograms measured from a subject and a predetermined set of microstate templates, An information processing device comprising means for reducing the state transitions between the microstates to a predetermined number of transition components by clustering them, wherein each of the predetermined number of transition components includes a portion of the transition pairs of a plurality of types defined as state transitions.

Images