Evaluation method, evaluation program, and information processing device

Abstract

To provide a method for evaluating a noise removal from a brain wave measured by a subject in an environment where an artificial magnetic field is present.SOLUTION: An evaluation method includes steps of: calculating a first retest reliability between a first brain wave and a third brain wave; calculating a second retest reliability between the first brain wave and a second brain wave; calculating a third retest reliability between the third brain wave and a fourth brain wave; calculating a fourth retest reliability between the second brain wave and the fourth brain wave; and outputting an evaluation result that a noise removal from the second brain wave and the fourth brain wave is appropriate when the fourth retest reliability satisfies a predefined standard. The second brain wave and the fourth brain wave are generated by removing noise from a signal measured from a subject.SELECTED DRAWING: Figure 2

Description

[Technical Field]

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

[0002] One known method for analyzing brain activity is to simultaneously perform electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) (hereinafter referred to as "EEG-fMRI simultaneous measurement").

[0003] Brain waves measured by EEG (electroencephalogram or electroencephalogram) have high temporal resolution, while fMRI measurements have high spatial resolution. Simultaneous EEG-fMRI measurement makes it possible to analyze brain activity by utilizing the advantages of both EEG and fMRI.

[0004] For example, by predicting fMRI signals, which are delayed by a few seconds, based on EEG signals, which immediately indicate neural activity, it is possible to effectively implement immediate neurofeedback. Furthermore, by accumulating the results of simultaneous EEG-fMRI measurements, it will eventually be possible to eliminate the need for fMRI, thereby realizing practical and significantly reduced-cost neurofeedback.

[0005] In simultaneous EEG-fMRI measurements, the brain waves measured by EEG can be contaminated with complex and large noise caused by fMRI. Examples of noise include radio waves used in MRI imaging and induced currents generated by physical changes in the EEG electrodes (due to body movement, heartbeat, etc.) in a high magnetic field. Therefore, methods have been proposed to remove noise from the brain waves measured by EEG. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] Antonova, E., Holding, M., Suen, HC, Sumich, A., Maex, R., & Nehaniv, C. (2022). EEG microstates: Functional significance and short-term test-retest reliability. NeuroImage: Reports, 2(2), 100089. https: / / doi.org / 10.1016 / j.ynirp.2022.100089 [Non-patent document 2] Khanna, A., Pascual-Leone, A., & Farzan, F. (2014). Reliability of resting-state microstate features in electroencephalography. PloS one, 9(12), e114163. https: / / doi.org / 10.1371 / journal.pone.0114163 [Non-patent document 3] Kleinert, T., Koenig, T., Nash, K., & Wascher, E. (2023). On the reliability of the EEG microstate approach. Brain Topography, 10.1007 / s10548-023-00982-9. https: / / doi.org / 10.1007 / s10548-023-00982-9 [Non-patent document 4] Liu, J., Xu, J., Zou, G., He, Y., Zou, Q., & Gao, JH (2020). Reliability and individual specificity of EEG microstate characteristics. Brain Topography, 33(4), 438-449. https: / / doi.org / 10.1007 / s10548-020-00777-2 Summary of the Invention [Problem to be solved by the invention]

[0007] In simultaneous EEG-fMRI measurements, it is physically impossible to obtain the true values ​​of the brain waves measured by EEG. Since true values ​​cannot be obtained, it is difficult to evaluate the noise removal method itself. As a result, there is a risk of abuse, such as arbitrary noise removal.

[0008] The present invention provides a method for evaluating noise reduction for electroencephalograms measured from a subject placed in an environment where an artificial magnetic field is present. [Means for solving the problem]

[0009] (Configuration 1) An evaluation method according to one embodiment of the present invention includes the steps of acquiring a first electroencephalogram (EEG) measured at a first time period, the first EEG indicating the electroencephalogram of a subject placed in an environment free of an artificial magnetic field, a second EEG indicating the electroencephalogram of the subject placed in an artificial magnetic field, a third EEG measured at a second time period, the second EEG indicating the electroencephalogram of the subject placed in an environment free of an artificial magnetic field, and a fourth EEG indicating the electroencephalogram of the subject placed in an artificial magnetic field. The second EEG and the fourth EEG are generated by performing noise removal on signals measured from the subject. The evaluation method includes the steps of: calculating a first retest reliability between the first EEG and the third EEG; calculating a second retest reliability between the first EEG and the second EEG if the first retest reliability satisfies a predetermined criterion; calculating a third retest reliability between the third EEG and the fourth EEG if the first retest reliability satisfies the predetermined criterion; calculating a fourth retest reliability between the second EEG and the fourth EEG if the second retest reliability and the third retest reliability satisfy the predetermined criterion; and outputting an evaluation result indicating that noise removal for the second EEG and the fourth EEG is appropriate if the fourth retest reliability satisfies the predetermined criterion.

[0010] (Configuration 2) In configuration 1, each of the first retest reliability, the second retest reliability, and the third retest reliability may be calculated based on an index of a microstate estimated based on the corresponding electroencephalogram.

[0011] (Configuration 3) In configuration 1 or 2, each of the first retest reliability, the second retest reliability, and the third retest reliability may be an intraclass correlation coefficient calculated for the microstate index.

[0012] (Configuration 4) In any of configurations 1 to 3, the step of calculating the second retest reliability may be performed for each of a plurality of second electroencephalograms generated by varying a plurality of parameters for noise removal. The step of calculating the third retest reliability may be performed for each of a plurality of fourth electroencephalograms generated by varying a plurality of parameters for noise removal.

[0013] (Configuration 5) In any of configurations 1 to 4, the evaluation method may further include a step of determining, as a first candidate parameter, a parameter for noise removal whose second retest reliability satisfies a predetermined criterion, and a step of determining, as a second candidate parameter, a parameter for noise removal whose third retest reliability satisfies a predetermined criterion. The step of calculating the fourth retest reliability may include a step of calculating a fourth retest reliability between a second electroencephalogram generated in accordance with the first candidate parameter and a fourth electroencephalogram generated in accordance with the second candidate parameter.

[0014] (Configuration 6) In any of configurations 1 to 5, the first, second, third, and fourth electroencephalograms may be generated by performing second noise removal on signals measured from the subject. The step of calculating the first retest reliability may be performed for each of the plurality of first electroencephalograms and the plurality of third electroencephalograms generated by varying a second parameter for the second noise removal. The evaluation method may further include the step of determining the second parameter based on the calculated plurality of first retest reliabilities.

[0015] (Configuration 7) In any of configurations 1 to 6, the evaluation method may further include a step of outputting an evaluation result indicating that noise removal for the second EEG and the fourth EEG is inappropriate if any of the first retest reliability, the second retest reliability, the third retest reliability, and the fourth retest reliability does not satisfy a predetermined standard.

[0016] (Configuration 8) There is provided an evaluation program for causing a computer to execute the evaluation method according to any one of configurations 1 to 7.

[0017] (Configuration 9) An information processing device according to an embodiment of the present invention includes a memory unit that stores a first electroencephalogram (EEG) measured at a first time period, the first EEG indicating the electroencephalogram of a subject placed in an environment free of an artificial magnetic field, a second EEG indicating the electroencephalogram of a subject placed in an artificial magnetic field, a third EEG measured at a second time period, the second EEG indicating the electroencephalogram of a subject placed in an environment free of an artificial magnetic field, and a fourth EEG indicating the electroencephalogram of a subject placed in an artificial magnetic field, and a processing unit. The second EEG and the fourth EEG are generated by performing noise removal on signals measured from the subject. The processing unit is configured to execute the following processes: calculating a first retest reliability between the first EEG and the third EEG; calculating a second retest reliability between the first EEG and the second EEG if the first retest reliability satisfies a predetermined criterion; calculating a third retest reliability between the third EEG and the fourth EEG if the first retest reliability satisfies the predetermined criterion; calculating a fourth retest reliability between the second EEG and the fourth EEG if the second retest reliability and the third retest reliability satisfy the predetermined criterion; and outputting an evaluation result indicating that noise removal for the second EEG and the fourth EEG is appropriate if the fourth retest reliability satisfies the predetermined criterion. [Effects of the Invention]

[0018] According to the present invention, it is possible to evaluate noise removal for electroencephalograms measured from a subject placed in an environment where an artificial magnetic field is present. [Brief explanation of the drawings]

[0019] [Figure 1] FIG. 1 is a schematic diagram showing an example of the configuration for simultaneous EEG-fMRI measurement. [Figure 2] FIG. 1 is a schematic diagram showing an overview of an evaluation method according to an embodiment of the present invention. [Figure 3] FIG. 10 is a diagram for explaining an example of preprocessing in the evaluation method according to the present embodiment. [Figure 4]FIG. 10 is a diagram for explaining a process for calculating test-retest reliability between independently measured electroencephalograms. [Figure 5] FIG. 10 is a diagram showing an example of the calculation results of test-retest reliability between independently measured electroencephalograms and simultaneously measured electroencephalograms. [Figure 6] FIG. 10 is a diagram showing an example of calculation results of test-retest reliability between simultaneously measured electroencephalograms. [Figure 7] FIG. 1 is a schematic diagram illustrating an example of a hardware configuration of an information processing device according to an embodiment of the present invention. [Figure 8] 10 is a flowchart showing an example of a processing procedure of the evaluation method according to the present embodiment. [Figure 9] 10 is a flowchart showing an example of a processing procedure of the evaluation method according to the present embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of signal strength by frequency. DETAILED DESCRIPTION OF THE INVENTION

[0020] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail with reference to the accompanying drawings, in which the same or corresponding parts are designated by the same reference numerals and will not be described repeatedly.

[0021] [A. EEG-fMRI simultaneous measurement] First, we will explain the simultaneous EEG-fMRI measurement.

[0022] 1 is a schematic diagram showing an example of the configuration of EEG-fMRI simultaneous measurement. Referring to FIG. 1, a measurement system 1 includes an information processing device 100, an EEG measurement device 200, and an fMRI measurement device 300.

[0023] The information processing device 100 acquires an EEG signal measured by an EEG measurement device 200. The information processing device 100 may acquire an fMRI signal (functional magnetic resonance image) measured by an fMRI measurement device 300.

[0024] The EEG measurement device 200 measures EEG signals (electrical signals) indicating electroencephalograms from an EEG cap 2 worn on the head of a subject. The EEG measurement device 200 includes a multiplexer 202, a noise filter 204, an A / D (Analog to Digital) converter 206, a storage unit 208, and an interface 210.

[0025] The multiplexer 202 sequentially selects pairs of electrodes (channels) from the group of electrodes on the EEG cap and electrically connects them to the noise filter 204. The noise filter 204 includes a high-frequency cut filter and the like, and removes noise components contained in the EEG signal that indicates the brain waves occurring in the selected channel.

[0026] The A / D converter 206 samples the electrical signal (analog signal) output from the noise filter 204 at predetermined intervals and outputs the sampled signal as a digital signal.

[0027] The storage unit 208 sequentially stores the time-series data (digital signal) output from the A / D converter 206 in association with information indicating the selected channel and timing (for example, time or counter value).

[0028] The interface 210 outputs the time-series data of the EEG signal stored in the storage unit 208 to the information processing device 100 in response to an access from the information processing device 100 or the like.

[0029] The information processing device 100 performs noise removal and evaluation of the noise removal results, as will be described later.

[0030] [B. Overview of evaluation method] Next, an overview of the evaluation method according to this embodiment will be described.

[0031] Fig. 2 is a schematic diagram showing an overview of an evaluation method according to the present embodiment. Referring to Fig. 2, in this embodiment, two measurement results taken at different times are used. The first measurement (measurement at a first time point) and the second measurement (measurement at a second time point) may be separated by, for example, one or more days. In other words, it is preferable that the first measurement and the second measurement are taken on different days.

[0032] In each session, EEG-only measurement (without fMRI measurement) and EEG-fMRI simultaneous measurement are performed on the same subject. Since EEG-only measurement and EEG-fMRI simultaneous measurement cannot be performed simultaneously, they are performed at different times. However, the time difference between the EEG-only measurement and the EEG-fMRI simultaneous measurement is sufficiently short compared to the time difference between the first and second measurements.

[0033] Hereinafter, electroencephalograms acquired by EEG-only measurement will also be referred to as "single-measurement electroencephalograms," and electroencephalograms acquired by EEG-fMRI simultaneous measurement will also be referred to as "simultaneous measurement electroencephalograms."

[0034] In EEG-only measurement, the subject is not exposed to a magnetic field generated by the fMRI measurement device 300. Therefore, the EEG-only measured brain waves correspond to brain waves of the subject placed in an environment free from any artificial magnetic field.

[0035] In the EEG-fMRI simultaneous measurement, the subject is exposed to a magnetic field generated by the fMRI measurement device 300. Therefore, the simultaneously measured electroencephalogram corresponds to an electroencephalogram representing the electroencephalogram of the subject placed under an artificial magnetic field. In this specification, the term "data set" refers to a set of an independently measured electroencephalogram 34-1 (corresponding to the first electroencephalogram) and a simultaneously measured electroencephalogram 36-1 (corresponding to the second electroencephalogram) obtained by a first measurement, and an independently measured electroencephalogram 34-2 (corresponding to the third electroencephalogram) and a simultaneously measured electroencephalogram 36-2 (corresponding to the fourth electroencephalogram) obtained by a second measurement.

[0036] In the evaluation method according to the present embodiment, preprocessing including noise removal is performed on the EEG signals (raw data) of each electroencephalogram included in the data set.

[0037] First, the test-retest reliability between the first independently measured electroencephalogram 34-1 and the second independently measured electroencephalogram 34-2 is calculated (step S1).

[0038] Next, the test-retest reliability between the first independently measured electroencephalogram 34-1 and the first simultaneously measured electroencephalogram 36-1 is calculated (step S2). Similarly, the test-retest reliability between the second independently measured electroencephalogram 34-2 and the second simultaneously measured electroencephalogram 36-2 is calculated (step S2).

[0039] Furthermore, the test-retest reliability between the first simultaneously measured electroencephalogram 36-1 and the second simultaneously measured electroencephalogram 36-2 is calculated (step S3). A higher test-retest reliability calculated in step S3 indicates a higher evaluation of the noise removal result.

[0040] The noise removal algorithm or parameters employed in the preprocessing may be changed or adjusted based on the evaluation results in step S3, or the noise removal algorithm itself may be evaluated based on the evaluation results in step S3.

[0041] Test-retest reliability refers to the degree of stability or consistency of measurement results over a certain period of time. Test-retest reliability indicates the reliability of a measurement between measurements taken on two occasions (the first and second times). As an example, a processing example will be described in which intraclass correlation coefficients (hereinafter also abbreviated as "ICC") are used as test-retest reliability.

[0042] Test-retest reliability may be calculated based on EEG microstates, which have been reported to have high test-retest reliability when EEG is measured alone (see Non-Patent Documents 1 to 4). In this way, test-retest reliability may be calculated based on microstate indices estimated based on corresponding EEGs (single-measurement EEGs and simultaneous measurement EEGs). A microstate refers to a typical spatial pattern extracted by clustering an EEG topology map. The number of microstates is typically four, but may be four or more. Microstate indices (such as duration, occurrence, and coverage) contain time information from EEGs, allowing evaluation of the effectiveness of noise reduction in brain activity dynamics.

[0043] Thus, test-retest reliability may be the intraclass correlation coefficient (ICC) calculated for the microstate index.

[0044] Note that test-retest reliability may be calculated using another index, not depending on EEG microstates.

[0045] The evaluation method according to this embodiment is based on the premise that if the retest reliability between the first individually measured EEG 34-1 and the second individually measured EEG 34-2 is high, then the retest reliability between the individually measured EEG 34 and the simultaneously measured EEG 36 measured on the same day should also be equally high, but that the retest reliability does not increase due to noise being mixed into the simultaneously measured EEG 36. In addition, the evaluation method according to this embodiment is based on the new finding of removing noise so as to increase the retest reliability between the individually measured EEG 34 and the simultaneously measured EEG 36 measured on the same day.

[0046] [C. Details of evaluation method] Next, the evaluation method according to this embodiment will be described in detail.

[0047] (c1: Preprocessing of EEG signals) The acquired EEG signals (raw data) are subjected to preprocessing, for example, as follows: The preprocessing includes a process for removing noise.

[0048] <Single-measurement EEG> (1) High-pass filtering (e.g., 0.5 Hz) (3) Low-pass filtering (e.g., 125 Hz) (6) Resampling process (e.g., 250Hz) <Simultaneous EEG Measurement> (1) High-pass filtering (e.g., 0.5 Hz) (2) GA (Gradient Artifact) correction using AAS (Average Artifact Subtraction) (3) Low-pass filtering (e.g., 125 Hz) (4) Resampling process (e.g., 500Hz) (5) Reduction of motion-related artifacts by CWL regression (6) Resampling process (e.g., 250Hz) (7) Application of a Joint Decorrelation (JD)-based method with varying thresholds to reduce residual BCG (ballistocardiogram) <Simultaneous and single-measurement EEG> (8) Bandpass filtering (e.g., 2-20Hz) (9) Artifact removal by spectral thresholding (10) Set the re-reference data to a common average value (11) Exclude noisy channels (12)Independent Component Analysis (ICA) (13) Assigning independent component labels (ICLabels) using the same threshold set for the same artifact components between single- and simultaneous-measurement EEGs In this way, the simultaneously measured electroencephalogram 36 is generated by performing noise removal ((2), (5), (7), etc.) on the EEG signals (raw data) measured from the subject. The independently measured electroencephalogram 34 and the simultaneously measured electroencephalogram 36 are generated by performing other noise removal ((9), (11), (13), etc.) on the EEG signals (raw data) measured from the subject.

[0049] Fig. 3 is a diagram for explaining an example of preprocessing in the evaluation method according to the present embodiment. Referring to Fig. 3, preprocessing is performed on raw data 30 of the individually measured electroencephalogram and raw data 32 of the simultaneously measured electroencephalogram measured from the subject. As a result, preprocessed individually measured electroencephalogram 34 is generated from raw data 30 of the individually measured electroencephalogram, and preprocessed simultaneously measured electroencephalogram 36 is generated from raw data 32 of the simultaneously measured electroencephalogram.

[0050] In preprocessing of the simultaneously measured EEG, multiple different thresholds (BCG thresholds) for reducing residual BCG (hereinafter also abbreviated as "BCG thresholds") may be used. The BCG thresholds are an example of parameters for noise removal to reduce residual BCG.

[0051] In preprocessing of the independently measured EEG and the simultaneously measured EEG, multiple different thresholds for assigning ICLabel (ICLabel Thresholds) (hereinafter also abbreviated as "ICLabel thresholds") may be used. The ICLabel thresholds are an example of parameters (second parameters) for processing to estimate and remove noise characteristics (second noise removal).

[0052] By varying the ICLabel threshold and the BCG threshold, a plurality of pairs of individually measured EEG 34 and preprocessed simultaneously measured EEG 36 are generated from a set of raw data 30 of individually measured EEG and raw data 32 of simultaneously measured EEG. The pair of individually measured EEG 34 and preprocessed simultaneously measured EEG 36 corresponds to the first measured EEG in the data set.

[0053] Note that some of the noise removal processes, such as (5) reduction of motion-related artifacts using CWL regression, (7) application of a JD-based method with varying thresholds to reduce residual BCG, (9) artifact removal using spectral thresholding, and (13) assignment of independent component labels (ICLabels) using the same threshold set for the same artifact components between the single-measurement EEG and the simultaneous-measurement EEG, may be omitted, or additional noise removal processes may be added.

[0054] The ICLabel threshold may be optimized in step S1. Therefore, the process of generating the independently measured electroencephalogram 34 and the simultaneously measured electroencephalogram 36 with different ICLabel thresholds may be executed in step S1. That is, in step S1, the ICLabel threshold may be changed and the same process as the pre-processing may be executed again.

[0055] The process of generating the simultaneously measured electroencephalograms 36 with different BCG thresholds may be executed in step S2. That is, in step S2, the BCG thresholds may be changed and the same process as the pre-processing may be executed again.

[0056] In the following description, electroencephalograms with different ICLabel thresholds and / or different BCG thresholds are also referred to as "variations."

[0057] (c2: Test-retest reliability between independently measured EEGs (Step S1)) Step S1 is a process for evaluating whether or not the target data set can be used for the processes in step S2 and subsequent steps.

[0058] 4 is a diagram for explaining the process of calculating the test-retest reliability between independently measured electroencephalograms. Referring to FIG. 4, based on the first independently measured electroencephalogram 34-1 and the second independently measured electroencephalogram 34-2 included in each data set, the microstate 11 and the microstate 12 are estimated, and the microstate index is calculated.

[0059] A predetermined number (e.g., four) of microstate templates are estimated by clustering the topology map of the acquired EEG. Based on the microstate templates, it is estimated which microstate the subject is in at each time. Note that there are also times when the subject does not belong to any microstate.

[0060] The numbers associated with the letters (A, B, C, D) representing the microstates in Figure 4 indicate the spatial correlation coefficient of each microstate template between the first and second measurements. The number on the left indicates the average value of the four spatial correlation coefficients. The spatial correlation coefficient indicates the degree of similarity in the shape of the microstates.

[0061] Duration means the average duration of each microstate. Occurrence means the average number of occurrences of each microstate per second. Coverage means the ratio of the duration of each microstate to the total duration of all microstates. Transition means the probability of transitioning from each microstate to other microstates.

[0062] For each of these four microstate indices (Duration, Occurrence, Coverage, Transition), the ICC (ICC calculation result 20) is calculated between the first (Day 1) single-measured EEG 34-1 and the second (Day 2) single-measured EEG 34-2.

[0063] The ICC calculation result 20 includes an ICC for each microstate index for each microstate. In the ICC calculation result 20, the transition ICC is calculated for the transition probability from the corresponding microstate to each other microstate. For example, for microstate A, from the left, the ICCs for the transition probability from microstate A to microstate B, the transition probability from microstate A to microstate C, and the transition probability from microstate A to microstate D are shown. Similarly, for other microstates, the transition destination microstates are listed in the order A, B, C, D, etc.

[0064] The mean of each microstate index is calculated by Z-transforming the index value of each microstate, averaging the Z-transformed values, and then inversely transforming the averaged values ​​to return them to the original domain.

[0065] In the ICC calculation result 20 shown in Figure 4, the average values ​​of the microstate indices for Duration, Occurrence, and Coverage are 0.782, 0.730, and 0.677, respectively, indicating high test-retest reliability. As shown in Figure 4, a dataset that exhibits high test-retest reliability is used for processing in step S2 and subsequent steps. Note that in step S1, a dataset that does not exhibit high test-retest reliability may be excluded from processing in step S2 and subsequent steps.

[0066] (c3: Test-retest reliability between single-measurement EEG and simultaneous-measurement EEG (step S2)) Step S2 is for evaluating the results of noise removal for the raw data 32 of the simultaneously measured electroencephalograms for each time included in the target data set.

[0067] More specifically, similar to step S1, the microstate is estimated based on the first (or second) independently measured electroencephalogram 34 and the simultaneously measured electroencephalogram 36 of the target data set, and the microstate index is calculated.

[0068] For each of the calculated microstate indices, the ICC is calculated between the independently measured electroencephalogram 34 and the simultaneously measured electroencephalogram 36 .

[0069] As described above, variations with different BCG thresholds can be prepared for the simultaneously measured electroencephalogram 36, and therefore, ICC may be calculated for each BCG threshold.

[0070] 5 is a diagram showing an example of the calculation results of test-retest reliability between independently measured electroencephalograms and simultaneously measured electroencephalograms. Referring to FIG. 5, the ICC of the microstate index is calculated based on the independently measured electroencephalograms 34-1 and simultaneously measured electroencephalograms 36-1 from the first test (Day 1) included in a certain data set. The ICC of the microstate index is also calculated based on the independently measured electroencephalograms 34-2 and simultaneously measured electroencephalograms 36-2 from the second test (Day 2) included in a certain data set.

[0071] The horizontal axis of each graph in Fig. 5 is the BCG threshold. As an example, Fig. 5 shows the change in ICC when the BCG threshold is changed in increments of 0.01 in the range of 1.05 to 1.20. Note that "None" means that the JD-based method for reducing residual BCG is not applied.

[0072] For example, an ICC of 0.5 or greater may be considered to be high test-retest reliability, and therefore, if there is a BCG threshold for all microstate indices with an ICC of 0.5 or greater, the data set may be determined to exhibit sufficient test-retest reliability.

[0073] Alternatively, a BCG threshold value that results in an ICC of 0.5 or more may be determined as the optimum parameter to be used for noise removal.

[0074] (c4: Test-retest reliability between simultaneously measured EEGs (Step S3)) Step S3 is for optimizing the adequacy and parameters of the denoising process performed on the data set of interest.

[0075] More specifically, similar to step S1, microstates are estimated based on the first simultaneously measured electroencephalogram 36-1 and the second simultaneously measured electroencephalogram 36-2 of the target data set, and microstate indices are calculated.

[0076] For each of the calculated microstate indices, the ICC is calculated between the simultaneously measured electroencephalogram 36-1 and the simultaneously measured electroencephalogram 36-2.

[0077] As described above, variations with different BCG thresholds can be prepared for the simultaneously measured electroencephalogram 36, and therefore, ICC may be calculated for each BCG threshold.

[0078] 6 is a diagram showing an example of the calculation result of test-retest reliability between simultaneously measured electroencephalograms. Referring to FIG. 6, ICC (ICC calculation result 20) is calculated between the first simultaneously measured electroencephalogram 36-1 (Day 1) and the second simultaneously measured electroencephalogram 36-2 (Day 2).

[0079] The horizontal axis of each graph in Figure 6 is the BCG threshold for the first (Day 1) simultaneously measured EEG signal 36-1. As an example, Figure 6 shows the change in ICC when the BCG threshold is set to seven different values ​​(1.09, 1.13, 1.14, 1.15, 1.16, 1.17, and 1.18).

[0080] The graphs with different line types shown in Figure 6 show the change in ICC for the simultaneously measured EEG 36-2 on the second day (Day 1) when the BCG threshold was set to three different values ​​(1.13, 1.15, 1.16).

[0081] That is, FIG. 6 shows the ICC between variations of simultaneously measured electroencephalograms 36-1 in which seven different BCG thresholds are set, and variations of simultaneously measured electroencephalograms 36-2 in which three different BCG thresholds are set.

[0082] For example, an ICC of 0.5 or greater may be considered to be high test-retest reliability, and therefore, if there is a BCG threshold for all microstate indices with an ICC of 0.5 or greater, the data set may be determined to exhibit sufficient test-retest reliability.

[0083] Alternatively, a BCG threshold value that results in an ICC of 0.5 or more may be determined as the optimum parameter to be used for noise removal.

[0084] (c5: Application) The above-described processing of steps S1 to S3 allows the validity of the adopted noise removal algorithm and parameters to be evaluated. That is, for a data set that has been evaluated as having high retest reliability by the processing of steps S1 to S3, it can be determined that the noise removal algorithm and parameters adopted to generate the data set are valid for the data set.

[0085] In the processing of steps S1 to S3 described above, the parameters (BCG threshold and / or ICLabel threshold) can be changed to calculate the ICC for each parameter. The parameter that gives the highest calculated ICC value may be determined as the optimal parameter. In other words, the parameters used for noise removal may be optimized by the processing of steps S1 to S3.

[0086] In the processes of steps S1 to S3 described above, the noise removal results can be objectively evaluated based on the ICC value, and therefore the ICC may be used to search for algorithms and parameters suitable for noise removal.

[0087] In the following description, the BCG threshold and ICLabel threshold will be described as examples of parameters for noise removal, but various parameters can be adopted depending on the noise removal algorithm.

[0088] [D. Processing equipment] 7 is a schematic diagram showing an example of a hardware configuration of information processing device 100 according to this embodiment. Information processing device 100 can typically employ a personal computer that follows a general-purpose architecture.

[0089] Referring to FIG. 7, the information processing device 100 includes, as its main components, one or more processors 102, a main memory unit 104, a control interface 106, a network interface 108, an input unit 110, a display unit 112, and a secondary memory unit 120.

[0090] The processor 102 is made up of arithmetic processing circuits such as a CPU (Central Processing Unit) and a GPU (Graphical Processing Unit), and executes the processes described below by executing, in a specified order, codes included in various programs stored in the secondary storage unit 120. Therefore, the processor 102 corresponds to a processing unit.

[0091] The main memory unit 104 is configured with a volatile storage device such as a dynamic random access memory (DRAM) or a static random access memory (SRAM), and stores program code executed by the processor 102 and data referenced during program execution.

[0092] The control interface 106 exchanges data with an interface 210 of the EEG measurement device 200 .

[0093] The network interface 108 exchanges data with an external device (for example, a data server device on the cloud).

[0094] The control interface 106 and the network interface 108 are configured with any communication components such as a wired local area network (LAN), a wireless LAN, a universal serial bus (USB), or Bluetooth (registered trademark).

[0095] The input unit 110 is configured with a mouse, a keyboard, or the like, and receives operations from the user.

[0096] The display unit 112 is configured with a display or the like, and notifies the user of various information relating to the execution status of processing in the information processing device 100 and operations.

[0097] The secondary storage unit 120 is configured with a nonvolatile storage device such as a hard disk or a solid state drive (SSD), and stores programs executed by the processor 102, data necessary for executing the programs, and the like.

[0098] The secondary storage unit 120 stores raw data 122 (raw data 30 of the independently measured electroencephalogram and raw data 32 of the simultaneously measured electroencephalogram) collected from the EEG measurement device 200, preprocessed independently measured electroencephalograms 34, preprocessed simultaneously measured electroencephalograms 36, a preprocessing program 124 for executing the preprocessing, an evaluation program 126 for executing the processing of steps S1 to S3, and evaluation results 128 that are the processing results of steps S1 to S3. The secondary storage unit 120 may also store system programs such as an OS (Operating System). The "evaluation program" of the present invention may include not only the evaluation program 126 but also the preprocessing program 124.

[0099] [E. Processing Procedure] 8 and 9 are flowcharts showing an example of the processing procedure of the evaluation method according to the present embodiment. Each step shown in Fig. 8 and 9 may be realized by one or more processors 102 of the information processing device 100 executing the preprocessing program 124 and the evaluation program 126.

[0100] 8 and 9, the information processing device 100 selects raw data (first time) of electroencephalograms (single-measurement electroencephalograms and simultaneous-measurement electroencephalograms) measured on one day and raw data (second time) of electroencephalograms (single-measurement electroencephalograms and simultaneous-measurement electroencephalograms) measured on another day for the same subject (step S10). The information processing device 100 preprocesses the raw signals of the single-measurement electroencephalograms using different ICLabel thresholds to generate variations of the single-measurement electroencephalograms for each of the first and second times (step S11). In this way, the information processing device 100 acquires the single-measurement electroencephalograms (first electroencephalograms) measured at a first time period and the single-measurement electroencephalograms (second electroencephalograms) measured at a second time period.

[0101] The information processing device 100 estimates a microstate template from each variation of the independently measured electroencephalogram (step S12). The information processing device 100 calculates the spatial correlation coefficient of the microstate template between the first and second measurements for each ICLabel threshold value (step S13). The information processing device 100 determines the ICLabel threshold value corresponding to the highest spatial correlation coefficient among the calculated spatial correlation coefficients as the parameter to be used (step S14).

[0102] In this way, the information processing device 100 calculates an ICC for each of a plurality of first individually measured electroencephalograms and a plurality of second individually measured electroencephalograms generated by varying the ICLabel threshold value (a parameter for noise removal). Then, the information processing device 100 determines the ICLabel threshold value based on the calculated ICCs.

[0103] The information processing device 100 calculates the test-retest reliability (first test-retest reliability) between the first individually measured electroencephalogram and the second individually measured electroencephalogram.

[0104] More specifically, the information processing device 100 estimates a microstate from the independently measured electroencephalogram corresponding to the IC Label threshold determined in step S14 (step S15), and calculates a microstate index from the estimated microstate (step S16).The information processing device 100 calculates an ICC from the calculated microstate index (step S17).

[0105] The information processing device 100 determines whether the calculated ICC satisfies a predetermined standard (for example, 0.5 or more) (step S18).

[0106] If the calculated ICC does not satisfy a predetermined standard (NO in step S18), the information processing device 100 outputs an evaluation result indicating that the noise removal is inappropriate (step S40).

[0107] If the calculated ICC satisfies a predetermined standard (YES in step S18), information processing device 100 executes the processes in step S20 and thereafter.

[0108] Note that the number of microstate templates to be estimated may be varied in step S12. In this case, in step S13, the optimal number of microstate templates can be determined by calculating the spatial correlation coefficient of the microstate templates between the first and second times for each number of microstate templates.

[0109] After performing step S12, the estimated microstate template may be evaluated for its similarity to microstate templates presented in previous studies.

[0110] After step S11 is performed, it may be evaluated whether or not there are any artifacts at the subject level in the independently measured electroencephalogram.

[0111] The processing of steps S10 to S18 described above corresponds to the processing (step S1) for calculating the test-retest reliability between independently measured electroencephalograms.

[0112] The information processing device 100 preprocesses the simultaneously measured electroencephalograms by varying the BCG threshold value, thereby generating variations of the simultaneously measured electroencephalograms for each of the first and second measurements (step S20). In this manner, the information processing device 100 acquires the simultaneously measured electroencephalograms measured at the first time period (second electroencephalogram) and the simultaneously measured electroencephalograms measured at the second time period (fourth electroencephalogram). At this time, the ICLabel threshold value determined at step S14 is used.

[0113] When the test-retest reliability between the independently measured electroencephalograms satisfies a predetermined standard, the information processing device 100 calculates the test-retest reliability (second test-retest reliability) between the first independently measured electroencephalogram and the first simultaneously measured electroencephalogram.

[0114] More specifically, the information processing device 100 estimates a microstate for each BCG threshold from variations in the first simultaneously measured electroencephalogram (step S21), and calculates a microstate index for each BCG threshold from the estimated microstate (step S22). The information processing device 100 calculates an ICC for each BCG threshold from the microstate index of the first independently measured electroencephalogram calculated in step S16 and the microstate index for each BCG threshold (step S23). In this way, the information processing device 100 calculates an ICC for each of multiple simultaneously measured electroencephalograms generated with multiple different noise removal parameters (BCG thresholds).

[0115] The information processing device 100 determines whether any of the calculated first ICCs satisfies a predetermined standard (for example, 0.5 or more) (step S24). If none of the calculated first ICCs satisfies the predetermined standard (NO in step S24), the information processing device 100 outputs an evaluation result indicating that the noise removal is inappropriate (step S40).

[0116] Similarly, when the test-retest reliability between the independently measured electroencephalograms satisfies a predetermined standard, the information processing device 100 calculates the test-retest reliability (third test-retest reliability) between the second independently measured electroencephalogram and the second simultaneously measured electroencephalogram.

[0117] More specifically, the information processing device 100 estimates a microstate for each BCG threshold from the variations of the second simultaneously measured electroencephalogram (step S25), and calculates a microstate index for each BCG threshold from the estimated microstate (step S26). The information processing device 100 calculates an ICC for each BCG threshold from the microstate index of the second independently measured electroencephalogram calculated in step S16 and the microstate index for each BCG threshold (step S27). In this way, the information processing device 100 calculates an ICC for each of multiple simultaneously measured electroencephalograms generated with multiple different noise removal parameters (BCG thresholds).

[0118] The information processing device 100 determines whether any of the calculated second ICCs satisfies a predetermined standard (step S28). If none of the calculated second ICCs satisfies the predetermined standard (NO in step S28), the information processing device 100 outputs an evaluation result indicating that the noise removal is inappropriate (step S40).

[0119] If any of the calculated second ICCs satisfies a predetermined criterion (YES in step S28), the information processing device 100 determines the BCG threshold corresponding to the ICC that satisfies the predetermined criterion as a candidate parameter (step S29). The BCG threshold determined as a candidate parameter is determined for each of the first ICC and the second ICC.

[0120] In this way, the information processing device 100 determines as a candidate parameter a BCG threshold (parameter) whose test-retest reliability (first ICC) between the first independently measured EEG and the first simultaneously measured EEG meets a predetermined standard, and also determines as a candidate parameter a BCG threshold (parameter) whose test-retest reliability (second ICC) between the second independently measured EEG and the second simultaneously measured EEG meets a predetermined standard.

[0121] The processes of steps S20 to S29 described above correspond to the process (step S2) of calculating the test-retest reliability of the independently measured electroencephalogram and the simultaneously measured electroencephalogram.

[0122] The information processing device 100 calculates the retest reliability (fourth retest reliability) between the first simultaneously measured EEG and the second simultaneously measured EEG when the retest reliability (second retest reliability) between the first independently measured EEG and the first simultaneously measured EEG meets a predetermined standard and the retest reliability (third retest reliability) between the second independently measured EEG and the second simultaneously measured EEG meets a predetermined standard.

[0123] More specifically, the information processing device 100 calculates an ICC for each combination of BCG thresholds based on the microstate indices corresponding to the BCG thresholds of the candidate parameters determined in step S29 among the variations of the first simultaneously measured EEG and the microstate indices corresponding to the BCG thresholds of the candidate parameters determined in step S29 among the variations of the second simultaneously measured EEG (step S30).

[0124] In this way, the information processing device 100 calculates the ICC (fourth test-retest reliability) between the first simultaneously measured EEG generated according to the BCG threshold value of the candidate parameter and the second simultaneously measured EEG generated according to the BCG threshold value of the candidate parameter.

[0125] The information processing device 100 determines whether any of the calculated ICCs satisfies a predetermined standard (e.g., 0.5 or more) (step S31). If none of the calculated ICCs satisfies the predetermined standard (NO in step S31), the information processing device 100 outputs an evaluation result indicating that the noise removal is inappropriate (step S40).

[0126] If any of the calculated ICCs satisfies a predetermined criterion (YES in step S31), the information processing device 100 determines the BCG threshold value corresponding to the ICC that satisfies the predetermined criterion as the final parameter (step S32).

[0127] Then, the information processing device 100 outputs an evaluation result that the noise removal is appropriate (step S41), and also outputs the determined ICLabel threshold value and BCG threshold value (step S42).

[0128] In this way, when the retest reliability between the first and second simultaneously measured EEGs satisfies a predetermined standard, the information processing device 100 outputs an evaluation result indicating that noise removal from the first and second simultaneously measured EEGs is appropriate.

[0129] [F. Summary] For example, if the true value of the electroencephalogram can be obtained through simulation, the noise contained in the measured electroencephalogram can be removed by some method, and the effectiveness of the noise removal can be evaluated based on how close the value approaches the true value.

[0130] In general, it is difficult to obtain true values ​​of EEG signals. However, when EEG is measured alone, noise sources such as MRI can be physically eliminated, so it is thought that the true values ​​are closer to the true values ​​than when EEG signals are measured simultaneously. In addition, typical noise can be estimated and eliminated from EEG signals obtained by EEG-only measurement. For example, by performing independent component analysis on the EEG signal, it is possible to estimate components that correspond to the characteristics of typical noise.

[0131] Fig. 10 is a diagram showing an example of signal strength by frequency. Fig. 10 also shows an example of the power spectral density (PSD) of a separately measured electroencephalogram. In Fig. 10, "separately measured electroencephalogram" indicates an EEG signal acquired by EEG-only measurement.

[0132] Figure 10 shows an example of the power spectral density when GA correction is applied ("GA"), an example of the power spectral density when GA correction and CWL regression correction are applied ("GA+CWL"), and an example of the power spectral density when GA correction, CWL regression correction, and independent component label correction are applied ("GA+CWL+ICLabel").

[0133] As shown in Figure 10, removing typical noises can bring the EEG closer to that of a single-measurement EEG, but removing other noises is not easy.

[0134] On the other hand, simultaneous EEG-fMRI recording is often required to analyze brain activity, and in this case, it is necessary to remove noise from the EEG signal obtained by simultaneous EEG-fMRI recording.

[0135] Previous studies have reported that the effectiveness of noise reduction can be evaluated based on how closely the EEG signal obtained by EEG-only measurement approaches the true EEG signal. This evaluation method implicitly assumes that each measurement result has test-retest reliability, but there has been little previous research evaluating this assumption. Furthermore, the power spectral density (PSD) shown in Figure 10 is used as the evaluation index. However, the PSD loses temporal information, making it impossible to evaluate whether the effectiveness of noise reduction extends to brain activity dynamics.

[0136] In response to the above-described conventional problems, the method for evaluating noise removal results according to the present embodiment uses data sets of individually measured EEGs and simultaneously measured EEGs obtained through multiple measurements to calculate test-retest reliability for each of the individually measured EEGs, the individually measured EEGs and simultaneously measured EEGs, and the simultaneously measured EEGs. By evaluating the test-retest reliability of such EEGs, it is possible to objectively evaluate the noise removal technique and results. Furthermore, the evaluation method according to the present embodiment calculates test-retest reliability using a microstate index including time information of EEGs, making it possible to evaluate whether the effectiveness of noise removal extends to brain activity dynamics.

[0137] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0138] 1 Measurement system, 2 EEG cap, 11,12 Microstate, 20 ICC calculation results, 30,32,122 Raw data, 34 Single-measurement EEG, 36 Simultaneous measurement EEG, 100 Information processing device, 102 Processor, 104 Main memory, 106 Control interface, 108 Network interface, 110 Input unit, 112 Display unit, 120 Secondary memory, 124 Preprocessing program, 126 Evaluation program, 128 Evaluation results, 200 EEG measurement device, 202 Multiplexer, 204 Noise filter, 206 A / D converter, 208 Memory unit, 210 Interface, 300 fMRI measurement device.

Claims

1. The method comprises a step of acquiring a first electroencephalogram measured at a first time period, the first electroencephalogram indicating the electroencephalogram of the subject placed in an environment free from an artificial magnetic field, a second electroencephalogram indicating the electroencephalogram of the subject placed in an artificial magnetic field, a third electroencephalogram measured at a second time period, the second electroencephalogram indicating the electroencephalogram of the subject placed in an environment free from an artificial magnetic field, and a fourth electroencephalogram indicating the electroencephalogram of the subject placed in an artificial magnetic field, the second electroencephalogram and the fourth electroencephalogram are generated by performing noise removal on signals measured from the subject, calculating a first test-retest reliability between the first electroencephalogram and the third electroencephalogram; calculating a second test-retest reliability between the first electroencephalogram and the second electroencephalogram when the first test-retest reliability satisfies a predetermined criterion; calculating a third test-retest reliability between the third electroencephalogram and the fourth electroencephalogram when the first test-retest reliability satisfies a predetermined criterion; calculating a fourth retest reliability between the second electroencephalogram and the fourth electroencephalogram when the second retest reliability satisfies a predetermined criterion and the third retest reliability satisfies a predetermined criterion; and if the fourth retest reliability satisfies a predetermined standard, outputting an evaluation result indicating that noise removal for the second electroencephalogram and the fourth electroencephalogram is appropriate.

2. 2. The evaluation method of claim 1, wherein each of the first retest reliability, the second retest reliability, and the third retest reliability is calculated based on an index of a microstate estimated based on a corresponding electroencephalogram.

3. 3. The evaluation method according to claim 2, wherein each of the first retest reliability, the second retest reliability, and the third retest reliability is an intraclass correlation coefficient calculated for the microstate index.

4. the step of calculating the second test-retest reliability is executed for each of the second electroencephalograms generated using a plurality of different parameters for the noise removal; 4. The evaluation method according to claim 1, wherein the step of calculating the third retest reliability is performed for each of the plurality of fourth electroencephalograms generated by varying a parameter for the noise removal.

5. determining, as first candidate parameters, parameters for the noise removal whose second test-retest reliability satisfies a predetermined criterion; determining, as second candidate parameters, parameters for the noise removal whose third test-retest reliability satisfies a predetermined criterion; 5. The evaluation method according to claim 4, wherein the step of calculating the fourth test-retest reliability includes the step of calculating the fourth test-retest reliability between the second electroencephalogram generated according to the first candidate parameter and the fourth electroencephalogram generated according to the second candidate parameter.

6. the first electroencephalogram, the second electroencephalogram, the third electroencephalogram, and the fourth electroencephalogram are generated by performing a second noise removal on a signal measured from the subject; the step of calculating the first retest reliability is executed for each of the first electroencephalograms and the third electroencephalograms generated by varying a second parameter for the second noise removal; The evaluation method according to any one of claims 1 to 3, further comprising the step of determining the second parameter based on the calculated first test-retest reliabilities.

7. The evaluation method according to any one of claims 1 to 3, further comprising a step of outputting an evaluation result indicating that noise removal for the second electroencephalogram and the fourth electroencephalogram is inappropriate if any of the first retest reliability, the second retest reliability, the third retest reliability, and the fourth retest reliability does not satisfy a predetermined criterion.

8. An evaluation program for causing a computer to execute the evaluation method according to any one of claims 1 to 7.

9. a memory unit that stores a first electroencephalogram measured at a first time period, the first electroencephalogram indicating the electroencephalogram of the subject placed in an environment free from an artificial magnetic field, a second electroencephalogram indicating the electroencephalogram of the subject placed under an artificial magnetic field, a third electroencephalogram measured at a second time period, the second electroencephalogram indicating the electroencephalogram of the subject placed under an artificial magnetic field, and a fourth electroencephalogram indicating the electroencephalogram of the subject placed under an artificial magnetic field; a processing unit; the second electroencephalogram and the fourth electroencephalogram are generated by performing noise removal on signals measured from the subject, The processing unit calculating a first test-retest reliability between the first electroencephalogram and the third electroencephalogram; calculating a second retest reliability between the first electroencephalogram and the second electroencephalogram when the first retest reliability satisfies a predetermined criterion; calculating a third retest reliability between the third electroencephalogram and the fourth electroencephalogram when the first retest reliability satisfies a predetermined criterion; calculating a fourth retest reliability between the second electroencephalogram and the fourth electroencephalogram when the second retest reliability satisfies a predetermined criterion and the third retest reliability satisfies a predetermined criterion; and, when the fourth retest reliability satisfies a predetermined standard, outputting an evaluation result indicating that noise removal for the second electroencephalogram and the fourth electroencephalogram is appropriate.

Images