Available via license: CC BY
Content may be subject to copyright.
fnsys-16-786200 February 21, 2022 Time: 13:59 # 1
METHODS
published: 25 February 2022
doi: 10.3389/fnsys.2022.786200
Edited by:
Dietmar Plenz,
National Institute of Mental Health
(NIH), United States
Reviewed by:
Frederic Von Wegner,
University of New South Wales,
Australia
Anthony Zanesco,
University of Miami, United States
*Correspondence:
Tomohisa Asai
asai@atr.jp
Received: 30 September 2021
Accepted: 04 February 2022
Published: 25 February 2022
Citation:
Asai T, Hamamoto T, Kashihara S
and Imamizu H (2022) Real-Time
Detection and Feedback of Canonical
Electroencephalogram Microstates:
Validating a Neurofeedback System
as a Function of Delay.
Front. Syst. Neurosci. 16:786200.
doi: 10.3389/fnsys.2022.786200
Real-Time Detection and Feedback
of Canonical Electroencephalogram
Microstates: Validating a
Neurofeedback System as a
Function of Delay
Tomohisa Asai1*, Takamasa Hamamoto1,2, Shiho Kashihara1and Hiroshi Imamizu1,3
1Cognitive Mechanisms Laboratories, Advanced Telecommunications Research Institute International (ATR), Kyoto, Japan,
2Graduate School of Frontier Biosciences, Osaka University, Osaka, Japan, 3Department of Psychology, Graduate School
of Humanities and Sociology, The University of Tokyo, Tokyo, Japan
Recent neurotechnology has developed various methods for neurofeedback (NF), in
which participants observe their own neural activity to be regulated in an ideal direction.
EEG-microstates (EEGms) are spatially featured states that can be regulated through
NF training, given that they have recently been indicated as biomarkers for some
disorders. The current study was conducted to develop an EEG-NF system for detecting
“canonical 4 EEGms” in real time. There are four representative EEG states, regardless
of the number of channels, preprocessing procedures, or participants. Accordingly,
our 10 Hz NF system was implemented to detect them (msA, B, C, and D) and
audio-visually inform participants of its detection. To validate the real-time effect of this
system on participants’ performance, the NF was intentionally delayed for participants
to prevent their cognitive control in learning. Our results suggest that the feedback
effect was observed only under the no-delay condition. The number of Hits increased
significantly from the baseline period and increased from the 1- or 20-s delay conditions.
In addition, when the Hits were compared among the msABCD, each cognitive or
perceptual function could be characterized, though the correspondence between each
microstate and psychological ability might not be that simple. For example, msD should
be generally task-positive and less affected by the inserted delay, whereas msC is more
delay-sensitive. In this study, we developed and validated a new EEGms-NF system as a
function of delay. Although the participants were naive to the inserted delay, the real-time
NF successfully increased their Hit performance, even within a single-day experiment,
although target specificity remains unclear. Future research should examine long-term
training effects using this NF system.
Keywords: neurofeedback, delay, EEG microstates, control, sense of agency
INTRODUCTION
Recent neurotechnology has developed various methods for neurofeedback (NF),
in which participants observe their own neural activity to be regulated in an ideal
direction. This neuromodulation through long-term training could improve participants’
cognitive performance or even some clinical traits for people with mental disorders
Frontiers in Systems Neuroscience | www.frontiersin.org 1February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 2
Asai et al. Developing EEGms-NF System
(see for review, Watanabe et al., 2017;Lubianiker et al., 2019).
These previous findings suggest our mental adaptability, where
the NF system serves to assist our innate ability to learn new
mental states. Since neural activity itself does not produce any
sensory feedback within the brain, the system can “externalize”
that activity to be controlled by the participants themselves.
This strategy has been technically developed in terms of
brain-computer-interface (BCI), where participants can control
the externalized “sensory feedback” of their neural activity
through training.
Regarding the implementation of the NF system, there
are two options: spatial or temporal priority over the neural
representations of the brain. The former is achieved by functional
magnetic resonance imaging (fMRI)-NF, while the latter is mainly
achieved by electroencephalogram (EEG)-NF. In terms of clinical
application, resting-state functional connectivity (FC) in fMRI
analysis can be a target of fMRI-NF (i.e., FC-NF), since FC
is assumed to be a biomarker for some mental disorders (e.g.,
Yamada et al., 2017). More recently, however, resting-state EEG
signals have also been actively used in line this this, taking
advantage of its high temporal resolution to complement the
disadvantages of fMRI, which has a resolution of only a few
seconds, making it difficult to provide immediate feedback. One
useful EEG measure in the context of NF is the EEG microstate.
EEG-microstates (EEGms) are spatially featured expressions
that can be regulated through NF training (Diaz Hernandez
et al., 2016) since recent studies have also suggested EEGms as
biomarkers (e.g., da Cruz et al., 2020;Murphy et al., 2020).
Accordingly, the current study was conducted to develop an
EEG-NF system for detecting “canonical EEGms” in real time
(see Figure 1 for an overview). Once simple resting-state EEG
data are collected from a sufficient participant population, group-
level common EEG states can be obtained. Previous studies
have repeatedly reported that there are only four representative
EEG states, regardless of the number of channels, preprocess
procedures, or participants (Khanna et al., 2015). Although the
optimal number of canonical states remains debatable (Seitzman
et al., 2017;Michel and Koenig, 2018), the four most agreed-upon
spatial patterns are often called msA, B, C, and D (Figure 2).
Therefore, in this study, our 10 Hz NF system was implemented
to detect all four msABCD and audio-visually inform participants
of its detection, unlike a previous NF system for a specific
microstate (Diaz Hernandez et al., 2016).
The EEGms are observed as quasi-stable potentials lasting
60–120 ms that represent whole-brain network activity (Michel
and Koenig, 2018). The suggested procedure for EEGms analysis
consists of two main stages: clustering and labeling. At the
clustering stage, the standard deviation (global field power: GFP)
among all electrodes is first calculated for each recording session.
The time points of the greatest strength in the neuroelectric field,
namely GFP peaks, are well-reasoned to best represent periods of
momentary stability in the voltage topography (Zanesco, 2020).
Consequently, it is preferable to select GFP peak points to achieve
a high signal-to-noise ratio. Every spatial pattern at the local
maximum of each GFP time series is accumulated as the GFP
peak dataset for a participant group as a whole. An unsupervised
learning algorithm, such as K-means, conducts clustering of
the GFP peak dataset into optimal classes. Finally, group-wise
common topographies are obtained as templates (i.e., centroid)
for each class (typically, msA, B, C, and D). After that, the
labeling procedure calculates the similarity between the templates
obtained and every time point (including the GFP peak points)
of the participants and labels one of them (e.g., msA) for all
data points in a so-called “winner-take-all” manner (see section
“Discussion”, Mishra et al., 2020). Since each topography should
last for a while in a millisecond order, the original EEG data
with multiple channels are now simply seen as a state transition
pattern among four states, such as C, D, B, A, D, etc. As a
result, the EEGms analysis typically reveals the frequency of the
state (“occurrence”), the lasting of the state (“duration”), and
transition patterns among the states (“transition probability”)
for each recording session or each participant. These depicted
features of EEGms can be compared among participant groups as
potential biomarkers (de Bock et al., 2020;Perrottelli et al., 2021).
The current study examined the real-time effect of the
developed NF system, which has been implemented to detect
canonical msABCD (group-wise common topographies) and
audio-visually inform participants of its detection. The audio-
visual NF, however, was intentionally delayed for participants
to prevent their control in learning (i.e., hit the targeted state;
see section “Materials and Methods”). We hypothesized that an
inserted delay would affect participants’ learning through the NF
system due to their inability to utilize the feedback. Although
EEG-NF is advantageous in terms of its temporal resolution in
comparison with fMRI-NF, the real-time effect has not been
examined well in the literature, especially for EEGms. Given
that the EEGms might be assumed as basic components of
consciousness, often referred to as “atoms of thoughts” (Koukkou
and Lehmann, 1987;Lehmann, 1992;Lehmann et al., 1998;
Changeux and Michel, 2006), and last for a short period of time,
EEGms-NF should become less effective with a certain delay
(e.g., over some hundreds of milliseconds). Aside from EEGms-
NF, several studies have attempted to implement real-time EEG
feedback (Zich et al., 2015;Pei et al., 2020). A recent study
examined the effect of latency in visual NF that falls within the
range of 300 to 1000 ms in terms of a parietal alpha rhythm
(Belinskaya et al., 2020), and concluded that the delay is a crucial
parameter that must be minimized to achieve the desired NF
effect. This policy is also motivation to enhance the “sense of
agency” for participants in their NF training: a feeling of “I
am the origin of the sensory feedback” (Gallagher, 2000). In
addition to many psychophysical studies that have demonstrated
that the feedback delay clearly reduces both participants’ sense of
agency (Asai and Tanno, 2007;Asai, 2016) and their performance
in motor control tasks (Tanaka et al., 2011), Evans et al. also
suggested that an inserted delay resulted in a reduced sense of
agency over the externalized feedback through a motor imagery-
based BCI system (Evans et al., 2015).
The aim of the current study was to develop a closed-loop
NF system and validate it as a function of the inserted delay.
Participants were instructed to attempt to make more “Hits”
with audio-visual feedback (see Figures 3,4for the definition
of “Hits”). However, in some conditions (i.e., “sessions” in
the current case) the additional delay was inserted between
Frontiers in Systems Neuroscience | www.frontiersin.org 2February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 3
Asai et al. Developing EEGms-NF System
FIGURE 1 | Schematics of EEGms neurofeedback system. The referenced bipolar signals are processed every 100 ms (10 Hz) into an epoch-averaged spatial
pattern. This current state is compared with four EEGms templates on the basis of spatial similarity (Pearson’s correlation). When the largest value among the four
absolute correlation coefficients is greater than 0.8 (for example) threshold, the display suggests a green circle in the corresponding area. The targeted state is
suggested at the same time using a blue frame. If the green circle overlaps the blue frame, the participants receive a ringing sound as a reward.
FIGURE 2 | Canonical EEGms templates as the targeted states in neurofeedback. The microstate maps obtained from external participants under the eyes-closed
resting state are shown. (A) Four maps are identified through a typical microstate analysis (upper) based on the GFP (global field power) peak dataset (lower). The
sequences of microstate classes were determined by back-fitting to the data with the highest topographical correlation (see text for details). (B) These canonical
templates were spatially congruent with previous studies, regardless of the EEG measurement or analysis tools. (C) Polarity-ignored spatial similarity (Pearson’s | r|)
with normative templates in which both configurations were interpolated onto 67 ×67 grids.
participants’ neural activity and its externalized feedback in a
secret manner (Figure 5 for experimental design). If real-time NF
is effective for learning, participants’ Hit performance should be
increased from the baseline period and from delayed conditions.
In particular, given that each EEGms could be a different
cognitive unit, as previously discussed, the controllability of each
microstate in the real-time NF situation and robustness of the
controllability to a feedback delay may be different depending
Frontiers in Systems Neuroscience | www.frontiersin.org 3February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 4
Asai et al. Developing EEGms-NF System
on its cognitive responsibility. Therefore, we also compared
the real-time effects of msA, B, C, and D in terms of each
cognitive functionality.
MATERIALS AND METHODS
Participants
A total of 18 young naive participants (9 females, mean
age = 26.3) were recruited from the local community and paid for
their participation. All participants reported normal or corrected-
to-normal vision and hearing. All participants provided written
informed consent before the experiments were conducted. The
experiment was conducted in accordance with the principles of
the Declaration of Helsinki. The protocol of the present study
was approved by the local ethics committee (reference number
is 20–144 for the Ethics Committee of ATR).
Preparing for EEG-Microstates
Templates
The templates used in our NF system to be matched with the
participants’ current state were prepared in our previous study
as a set of independent group-wise common topography, where
31 healthy people (15 females) in their 20 to 60 s were recorded
under the four resting-like eye-closed conditions (resting state,
meditated state, respiration counting, and heartbeat counting).
Although the details for that experiment will be published
elsewhere, the canonical four templates (msA, B, C, and D) were
obtained from all those data, as briefly described below.
The EEG signals were recorded from 32 silver-silver chloride
electrodes attached to a saltwater sponge-based electro-cap (R-
net, Brain Products GmbH, Germany) and were placed at Fp1,
Fp2, Fz, F3, F4, F7, F8, F9, F10, FC1, FC2, FC5, FC6, Cz, C3,
C4, T7, T8, CP1, CP2, CP5, CP6, Pz, P3, P4, P7, P8, P9, P10,
Oz, O1, and O2, according to the extended international 10–
20 Systems. The reference and ground electrodes were placed
at FCz and Fpz, respectively. We maintained impedances under
50 k. EEG signals were amplified with a bandpass of 0.016–
250 Hz and digitized at a 500 Hz sampling rate using an EEG
recorder (BrainAmp ExG, Brain Products GmbH, Germany).
The preprocessing of the EEG data was conducted in MATLAB
(R2019b, MathWorks, United States) using the EEGLAB toolbox
(EEGLAB2019_0; Delorme and Makeig, 2004). First, the raw
signals were down-sampled to 100 Hz and filtered using a finite
impulse response filter with a high pass of 2 Hz and a low-
pass filter of 20 Hz. Then, we conducted a visual inspection
to detect artifacts, including sweat, muscle, movement, and
electrode trouble, to be removed manually. Channels with
severe artifacts during the entire recording period were spatially
interpolated. Furthermore, independent component analysis
(ICA) was conducted to remove components with artifacts.
Microstate analysis was further applied to this clean dataset in
MATLAB R2019b using the MST plugin for EEGLAB (Poulsen
et al., 2018). We calculated the GFP and accumulated topographic
voltage maps at local maxima (peaks) in the GFP time series (1000
GFP peaks per session as a default setting) for all participants to
be analyzed by the modified k-means clustering algorithm, which
ignores the polarity of the voltage maps. GFP peaks were used to
generate initial maps for clustering to maximize the topographic
signal-to-noise ratio. We defined the number of microstates as
four, given that previous studies reported them as the most
common (Khanna et al., 2014, 2015), although recent studies
have also argued for the possibility of more canonical templates
(e.g., Wackermann et al., 1993;Seitzman et al., 2017;Michel and
Koenig, 2018).
Developing EEG-Microstates-NF System
The EEGms-NF system was developed using OpenViBE (Renard
et al., 2010) with embedded MATLAB code so that the same
EEG cap and amplifier work with our templates (Figure 1).
The input EEG signals (500 Hz) were first referenced and then
bandpass-filtered (FIR, 2–20 Hz). After this minimum online
denoising, our system implemented time-epoching, template-
matching, thresholding, and displaying at 10 Hz. The epoching
module in OpenViBE determines the time window that averages
the online EEG signals as the “current” EEG topology that
consists of a 32-dimensional vector, as a result. Since the
typical duration of each microstate has been reported to be
approximately 100 ms (Khanna et al., 2015; 60–120 ms, Koenig
et al., 2002;Michel and Koenig, 2018), our 10 Hz system depicts a
100-ms–averaged topology for every process (i.e., for sequential
100-ms blocks without an overlap). This means that the only
temporally stable microstate (duration of approximately 100 ms)
should be detected for the following process. In this sense,
a polarity reversal (e.g., A + to A−) should self-cancel the
topography and should not be detected by the neurofeedback
system at this stage.
Because our four prepared templates are also 32D vectors,
the template matching was simply applied by calculating the
spatial similarity (Pearson’s correlation in our case) between
two 32D vectors for each template (i.e., the current vs. each
template). As a result, the system output includes four time-series
of correlation coefficients that range from −1.0 to + 1.0 in the
definition. For the following, however, the absolute rvalue was
used because the polarity is not of interest regarding EEGms
(see above for definition of microstate). At this stage, we can
define the threshold parameter for the absolute rvalues (see
Figure 3A for our 10-Hz system). If the system is processed
at 100 Hz (Figure 3B shows as an offline processing for
comparison), we observe positive–negative fluctuation regardless
of the template (Figure 3B) when we define positive-negative
templates as anterior-posterior contrasts. Since our template
A may not be optimal (see Figure 2C diagonals), spatial or
temporal correlations between template A and B in the current
case could be exceptional. For participants to learn to effectively
“hit the target” by controlling their own neural state, audio-
visual feedback should be appropriate in terms of its frequency
(Figure 4). If the threshold is too low (e.g., rthr = 0.1), participants
receive feedback too often to learn (i.e., it is annoying, Figure 4B).
Therefore, our pilot tests for the participants from our research
group determined the threshold as approximately 0.8 and further
individually adjusted (see below) in an explorative manner to
reduce the frequency of feedback (Figure 4A). This also means
Frontiers in Systems Neuroscience | www.frontiersin.org 4February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 5
Asai et al. Developing EEGms-NF System
FIGURE 3 | Detected canonical msABCD templates. (A) The exemplified time series of four correlation coefficients with template msABCD at 10 Hz is shown. The
yellow circle indicates the detected canonical microstates in real-time (when rthr = 0.8). (B) If the system is processed at 100 Hz (for comparison), a continuous
state-transition dynamic can be observed. In this sense, our time-averaged 10-Hz system detects only spatio-temporally robust microstates.
that only the spatially robust microstate (topology near the
templates) should be detected for the following process.
Finally, when the current topology passed the above-
mentioned criteria for a spatio-temporally robust microstate, a
display for participants indicated its detection (Figure 1) where
participants always saw four frames with labels (A, B, C, or D).
Only when that state was detected, a green circle appeared within
the corresponding frame (the highest-r label is selected so that
only one green circle was shown at a certain moment). At the
same time, a blue frame was presented among four possible states
as the target. When participants’ EEGms detected by the system
hit the target (a green circle appears within a blue frame), a
reward sound was played. Therefore, in short, participants were
instructed to “generate more sounds” whereas the blue target
randomly moved every 20 s. There were five target conditions
(no target, A, B, C, and D) within a session of 5 min (20 s ×5
conditions ×3 repetitions) (Figure 5A).
Behavior of the System Developed
In summary, the developed system was implemented to detect
“four canonical microstates” in terms of the duration and
spatial pattern, where those definitions (e.g., threshold) were
parameterized. The current setting (time window = 100 ms and
spatial similarity >0.8 ∼0.7) was intended to detect only spatio-
temporally robust states, so that participants received roughly
1–3 instances of visual feedbacks per 1 s by default (mean ±SD:
2.7 ±0.7 times/s) (Figure 4A) because EEGms transition should
be an innate neural dynamic. Accordingly, only minimal online
denoising was required, because potentially noisy states (e.g., eye
blink, head motion, and other possible artifacts) must be ignored
in the definition. We also confirmed that such intentionally added
noise does not respond to our system regardless of msA, B, C, or
D in preceding pilot tests. This indicates that participants could
not use physical strategies; instead, they were encouraged to try
only mental strategies to “make sounds”.
A previous study implemented the EEGms-NF system (Diaz
Hernandez et al., 2016) in which the temporally weighted
contribution of msD during the recent 1 s was fed back auditorily
to participants. This policy was based on a typical microstate
analysis, since each data point was assigned to msA, B, C,
or D without polarity (often followed by temporal smoothing
to ignore rapid transitions), whose templates were depicted
individually. The current NF system, however, has a different
approach to promote its real-time status on the basis of our
simpler definition of the “state.” We need only a recent 100-ms
data if the state lasts approximately that duration. This means,
in turn, that a polarity reversal and a rapid mixture among
Frontiers in Systems Neuroscience | www.frontiersin.org 5February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 6
Asai et al. Developing EEGms-NF System
FIGURE 4 | Outputs of the neurofeedback system. (A) A typical scatter plot of a 5-min session (left) showing visually informed (circle feedback) data points only
(middle) and auditory-informed “Hits” only (right). (B) If the threshold is lowered (e.g., rthr = 0.1 for comparison), participants receive too much feedback (visual for the
middle panel and auditory for the right panel) to learn.
msABCD should self-cancel the topography and should not be
detected by the neurofeedback system. Therefore, our current
system is conservative where only spatio (r>0.8 approximately)-
temporally (100 ms approximately) stable states should be
detected at every 10 Hz epoching process. Other possibilities,
including mistiming, polarity reversal, mixture among msABCD,
and intrinsic/extrinsic noisy states, were ignored. In addition,
our audio-visual feedback achieves supervised learning for
participants who are aware of their current state independent
of the target state (group-wise common topography). The
difference between them can serve as a prediction error to
be minimized through learning. Real-time feedback would
contribute to the calculation of prediction errors with a valid
temporal correspondence between the current state and the
target. Therefore, we hypothesized that participants could learn
to make more Hits through real-time feedback of the system,
compared with delay-inserted conditions, as shown below.
Inserting Delay for System Validation
For that purpose, we additionally inserted an audio-visual
delaying hardware (SS-FBDL82, SPORTS SENSING Co., LTD.,
JAPAN) between the NF system and both the monitor and
speaker (Figure 5B). The inserted delay parameters were 0, 1 s,
or 20 s and were manipulated session-wise in a manner blinded
to the participants. They repeated six sessions in total (three
conditions ×two repetitions in random order). Accordingly,
participants had three delay conditions as a between-session
factor and five target conditions as a within-session factor.
Previous studies have suggested that approximately a 1-s delay
or epoching may be long enough to disturb or reduce the
effect of EEG-NF (Mulholland et al., 1979;Belinskaya et al.,
2020). Furthermore, a 20-s delay was intended to make a total
discrepancy between the current and target state since the target
moves every 20 s (for example, participants see the target msA
then try to make an msA state of their own, but the presented
visual feedback was for the previous target condition other than
msA). If the functionality of the targeted microstate was not
perceptual, but more cognitive, a 1-s or even 20-s delay might
be acceptable, but in that case (especially for a 20-s delay), this is
not likely a pure EEG-NF effect, but should include a task general
effect similar to the results of a cognitive workload (see section
“Discussion”).
Participants first received instructions regarding EEG
measurement and the necessity to remain immobile during
experiments and about microstates in terms of controlling their
occurrence. They were then instructed to make more Hits (a
green circle within a blue frame) by changing their conscious
states in an explorative manner for which success was also
indicated by auditory ringing feedback. First, a 1-min practice
session was conducted using the same system without additional
delay, where the experimenters visually checked the response of
the system (i.e., audio-visual feedback) to determine individual
thresholds among 0.8, 0.75, and 0.7, to absorb potential
individual differences. Because the system responsiveness could
Frontiers in Systems Neuroscience | www.frontiersin.org 6February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 7
Asai et al. Developing EEGms-NF System
FIGURE 5 | Within- and between-session conditions. (A) A session consists of five within-session target conditions ×3 repetitions. (B) Participants completed 6
sessions (3 between-session delay conditions ×2 repetitions). The order of conditions was randomized.
be affected by several factors [a cap condition with potentially
bad channels or fit of the cap to the participant’s head (S, M, or
L size) as well as individual differences in participants’ innate
states in terms of the similarity with the canonical templates]. Six
participants for each threshold were determined as a result. Since
this was a simple manipulation to roughly equalize feedback
frequency among participants, the effect of the threshold would
not be examined (see also Discussion for further update of the
procedure). After the practice session, the participants repeated
six sessions with or without delay insertion, as mentioned above.
Participants were also required to answer a question regarding
subjective controllability of their own neural state toward each
EEGms (A, B, C, or D) as well as their sleepiness using a 5-point
Likert scale after each session. This was expected to be a proxy
measure for evaluating participants’ sense of agency over the
EEGms-NF. The total time for the current experiment per
subject was 2 h.
RESULTS
The templates we obtained are shown in Figure 2, where the
labels of maps were named according to previously established
studies (i.e., typically msA, B, C and D; Lehmann et al., 1987;
Koenig et al., 2002;Khanna et al., 2014, 2015;Diaz Hernandez
et al., 2016;Milz et al., 2016;Michel and Koenig, 2018). The
topography showing a left posterior to right anterior orientation
was determined as msA, the right posterior-left anterior
orientation was msB, the anterior-posterior orientation was msC,
and the fronto-central extreme location was msD. The spatial
configurations of these four maps were highly similar to those
previously described and frequently cited in EEG microstate
research (Koenig et al., 2002) (Figure 2B). Comparing each
two maps with the corresponding label assignment (Brodbeck
et al., 2012), we achieved high spatial correlation coefficients
(0.85–0.97, Figure 2C). Furthermore, we calculated the global
explained variance (GEV) of the templates. The GEV is one of
the parameters that can be used to evaluate whether this set
of maps is reasonable as a canonical template. It provides a
metric of how well the selected template maps account for the
variance of the entire dataset (Poulsen et al., 2018). The higher
the GEV, the better the explanation of the entire dataset. As a
result, our template of four cluster maps explained 75% of the
variance. Referring to previous studies (Figure 2B), the GEV
was 71% in Diaz Hernandez et al. (2016), who recorded and
preprocessed EEG signals with the same number of electrodes
and filter settings as ours, 70% in Khanna et al. (2014) (they used
a wider bandpass filter than ours), and 77% in Milz et al. (2016)
(they used a larger number of electrodes than ours). Therefore,
we obtained canonical microstates that were congruent with
previous observations. These templates were used in our NF
system to be matched with participants’ EEG states in real time.
Questionnaire and Raw Hit Scores
Participants’ raw reports about their subjective controllability
over EEGms based on their Hit performance are summarized
Frontiers in Systems Neuroscience | www.frontiersin.org 7February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 8
Asai et al. Developing EEGms-NF System
FIGURE 6 | Questionnaire ratings for controllability and the raw number of Hits. (A) Participants reported their subjective feelings regarding “controllability” over the
targeted EEGms and their sleepiness on the five-point Likert scale. (B) Participants’ raw Hits performance for each EEGms target. (C) A plot between the number of
raw Hits and the Likert rating, including the delay conditions. Error bars indicate ±SE.
TABLE 1 | Descriptive statistics for participants’ raw Hits and ratings.
(A) Raw hits (B) Ratings
Descriptive statistics Mean Std. deviation Minimum Maximum Descriptive statistics Mean Std. deviation Minimum Maximum
Total_0s 181.6 46.96 111 281 A_0s 3.194 0.957 1 5
Total_1s 169.6 49.74 93.5 268.5 B_0s 3.639 1.148 2 5
Total_20s 167.3 54.13 89.5 266 C_0s 2.222 1.396 1 5
msA_0s 43.39 31.25 17 160.5 D_0s 3.722 1.166 1.5 5
msA_1s 37.42 16.72 17 81 Sleep_0s 1.694 1.002 1 3.5
msA_20s 39.08 18.79 17 99.5 A_1s 3.25 1.018 1 5
msB_0s 28.81 11.68 15.5 61.5 B_1s 3.667 1.098 1.5 5
msB_1s 29 14.39 9 56.5 C_1s 2.25 1.364 1 5
msB_20s 30.17 12.65 9.5 56 D_1s 3.806 1.202 1.5 5
msC_0s 79.17 30.64 39 149 Sleep_1s 1.833 1.043 1 4
msC_1s 73.89 27.68 42.5 127.5 A_20s 3.139 1.026 1 5
msC_20s 67.89 29.08 37.5 136 B_20s 3.472 0.899 2 5
msD_0s 29.97 15.29 5 64 C_20s 2.194 1.33 1 5
msD_1s 29.14 13.87 2.5 54.5 D_20s 3.667 1.125 1.5 5
msD_20s 30 12.56 8 60.5 Sleep_20s 1.667 0.874 1 4
in Figure 6 and Table 1. The current results are congruent
with some previous studies in terms of the statistical features
of the four EEGms. For example, msC is reportedly the most
dominant at default, especially in young, healthy participants
(Tomescu et al., 2018). In line with this, the current participants’
raw number of Hits for msC was higher than that for msA, B,
or D, regardless of the inserted delay (Figure 6B). The two-way
ANOVA (4-target within conditions ×3-delay within condition)
revealed significant main effects of the target and an interaction
[F(3,51) = 24.2, p<0.0001, and F(6,102) = 2.36, p= 0.0355],
while the main effect of delay missed significance [F(2,34) = 24.2,
p= 0.0501]. Multiple comparisons with Ryan’s method for the
main effect of the target indicated that msC was significantly
increased compared to msA, B, or D, respectively (ps <0.05).
Frontiers in Systems Neuroscience | www.frontiersin.org 8February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 9
Asai et al. Developing EEGms-NF System
Regarding the interaction, the simple main effect of delay on
msC was significant [F(2,136) = 7.90, p= 0.0006]. The main
effect of delay on the total Hits was significant [one-way ANOVA,
F(2,34) = 3.36, p= 0.0466]. Accordingly, participants’ subjective
feelings about the difficulty in controllability were also reduced
for msC (Figure 6A). The two-way ANOVA (4-target within
conditions ×3-delay within condition) revealed a significant
main effect of target [F(3,51) = 7.93, p= 0.0002], while that of
delay or the interactions were not significant [F(2,34) = 0.49,
p= 0.6189, and F(6,102) = 0.09, p= 0.9969, respectively].
Multiple comparisons with Ryan’s method indicated that msC
was significantly reduced from msA, B, or D (ps <0.05).
In addition, sleepiness reports did not differ among 3-delay
conditions [one-way ANOVA, F(2,34) = 0.69, p= 0.5072].
This outcome-dependent relationship between NF or BCI
performance and perceived agency has been examined previously
(Evans et al., 2015;Caspar et al., 2021). In particular, when
the desired outcome is achieved, participants feel agency even
if that outcome is not achieved by their own neural activity
(Evans et al., 2015). Similarly, in the current study, it seems
that agency ratings might have simply been reflected on the
actual number of Hits by receiving or counting the sounds (e.g.,
for msC). Figure 6C clearly depicts the negative correlation
between them with collapsing four target conditions, regardless
of the inserted delay. However, regarding the effect of delay,
there was no difference in ratings for either controllability or
sleepiness, although Hit performance could be modulated by
the inserted delay (see below for details). This suggests, in
turn, that participants were not aware of the inserted delay,
unlike many behavioral studies, when even short-time delays are
easily detectable (Asai and Tanno, 2007). In such a situation,
participants’ perceived agency is correlated with the detection of
delay, which makes it difficult to eliminate the possibility that the
reduced agency also affects participants’ motivation to learn. In
this sense, no difference in their rating for the inserted delay (as
a manipulation check) should be an important controlling result,
especially in the current study.
The Relative Hits as a Function of Delay
To further examine the effect of delay on participants’ Hit
performance, the raw scores were individually standardized
based on the baseline period (no-target condition, see Figure 5A)
because raw Hits by default have individual differences and
also depend on the system threshold (rthr = 0.8, 0.75, or
0.7) that were not strictly determined individually in the
current study. Figure 7 indicates the relative Hits score, the
relative magnification in comparison to the individual no-
target condition without delay, as a function of the inserted
delay. Our results clearly suggest that the real-time effect was
observed only under the no-delay condition in the total score
(Figure 7A). The relative Hits significantly increased from
the 1- or 20-s delay conditions. A one-way ANOVA revealed
that the main effect of delay was significant [F(2,34) = 4.86,
p= 0.0139] with a difference between delay-0 and -1, and delay-0
and -20 revealed by a post hoc multiple comparisons using
Ryan’s method (ps <0.05). Furthermore, the two-way ANOVA
(4-target within conditions ×3-delay within condition) revealed
that the main effect of the target was significant [F(3,51) = 3.94,
p= 0.0128], while that of delay and interactions were not
[F(2,34) = 1.05, p= 0.3601, and F(6,102) = 1.04, p= 0.4032,
FIGURE 7 | Participants’ Hit performance as a function of delay. (A) The relative magnification of total Hits in comparison to the individual baseline (no-target
condition). (B) The same indices are calculated, respectively, for msA, B, C, and D. Error bars indicate ±1 SE. *p<0.05 in ANOVA.
Frontiers in Systems Neuroscience | www.frontiersin.org 9February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 10
Asai et al. Developing EEGms-NF System
FIGURE 8 | System outputs as an asymmetrical matrix. (A) The raw counts for an exemplified session (left) and the summarized matrix for the target specificity are
shown in contrast to possible baselines (right). (B) Actual participants’ averages as occupation ratios for each delay condition.
respectively]. The relative Hit for msD was significantly
increased from msA, B, and D, regardless of the delay, according
to a multiple comparisons using Ryan’s method (ps <0.05),
suggesting a potential difference in the functionality of msABCD
in terms of the responsiveness of our NF system with the inserted
delay (Figure 7B).
For the individual EEGms, the increase from the baseline
was the significant contrast of interest. For that purpose, we
applied a Bayesian one-sample t-test to the baseline to avoid
repetition of multiple tests (e.g., multiple Student’s t-tests). When
we took a two-sided alternative hypothesis (H1:δ6= 1, since the
relative Hits could be lower than 1) against the null hypothesis
(H0:δ= 1), a Bayesian one-sample t-test with JASP (JASP Team,
2019) revealed weak but interpretable evidence (BF10 >1) under
the no-delay condition for the total, msC, and msD (as well
as under the delay-1 or -20 condition for msD). For example,
BF10 = 1.52 for total score without delay, indicates that the data
are approximately 1.5 times more likely to occur under H1 than
under H0, suggesting at least weak evidence in favor of H1 (a
Bayes factor between 1 and 3 is considered weak evidence, but this
range might be common in behavioral sciences). A confirmatory
non-parametric t-test with Wilcoxon signed-rank suggested the
same results, where the five conditions with black-framed bars in
Figure 7 were significantly different from the baseline (ps <0.05).
Accordingly, when the Hits were compared among the msABCD,
each cognitive function could be characterized. The total result
is congruent solely with the msC, in which only the no-delay
condition increased significantly above the baseline. Although
a similar trend was detected for msA and msB, the difference
was not significant. The msD is especially interesting, given that
even a 20-s delay elevated the relative Hit performance compared
with baseline (no-target specified), suggesting its general task-
positive functionality. This in turn indicates the necessity of using
a task-general baseline (see below); a previous study suggested
the importance of using multiple baselines for assessing a training
effect (Alkoby et al., 2018).
Target Specificities for Each Delay
Condition
The output of the system is summarized in Figure 8, in which
a 4 ×4 (originally 5 ×5) matrix compares the Hits (diagonal
components) and Misses (non-diagonals). For the group average,
we assume that the “no-target” condition serves as a “task-
ready” state rather than a resting state (Figure 8B), since the
occupation ratio is almost that same as other “target” conditions
regardless of the delay. Indeed, participants could observe their
own EEG states as visual feedback even during the no-target
Frontiers in Systems Neuroscience | www.frontiersin.org 10 February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 11
Asai et al. Developing EEGms-NF System
period. Therefore, the individual baseline is needed to examine
a small effect from real-time feedback on learning (i.e., the effect
of a delay) within a single-day experiment. Since the definition
of the baseline might change the results, the “task-general”
baseline was also used for comparison with target conditions
(Figure 8A, right).
Figure 8 shows the target specificity analyzed with two
possible baselines: the no-target period (“task-ready”) under the
no-delay condition (similar to Figure 7) and the session average
(the column means for “task-general”) for each delay condition
(see Figure 8A, right). According to Figures 9A,B, the values
are generally reduced with a delay regardless of being a Hit
or Miss. A Hit for msC is the most sensitive to a delay, while
msA exhibits the same tendency as msC in Figure 9A but msB
shows the opposite in Figure 9B. The increase in msD evident
in Figure 9A almost disappears in Figure 9B. A three-way
ANOVA (4 targets ×4 currents ×3 delay conditions) applied
to the “task-general” baseline (Figure 9C) shows that the effects
of the delay and the target were significant [F(2,34) = 10.19,
p= 0.0003; F(3,51) = 5.501, p= 0.0024, respectively]; the
interaction between them was also significant [F(6,102) = 2.643,
p= 0.02]. Accordingly, the target specificity (i.e., the three-way
interaction or an interaction between the target and the detected
state) was not specifically confirmed, although the main effect of
a delay indicates a real-time effect from feedback, which was our
initial purpose.
Although the target specificity is still unclear, that effect can
be predicted to be learnable by participants through long-term
training using this feedback system in a future study. This is
related to their functional responsibility. Previous studies have
suggested that the functionality of msA and msB are perceptual
(auditory for msA and visual for msB); therefore, they might
be passive and less controllable (oriented for sensory inflow,
which is mainly audio–visual for humans), while msC and msD
are cognitive (saliency or default mode for msC and attentive
or executive for msD); therefore, they are active and more
controllable (oriented for motor outflow with switching from
internal to external awareness); below we discuss this functional
difference in terms of delay sensitivity in our EEGms-NF system.
DISCUSSION
In the current study, we developed a new EEGms-NF system and
validated it by examining the decline in participants’ performance
as a function of delay. Although participants were presumably
totally naive about the inserted delay as the result of the
agency questionnaire also implies (Figure 6A), the real-time NF
successfully increased their Hit performance in total (Figure 7A),
even within a single-day experiment. This suggests that by
using our NF system, participants could implicitly learn through
a supervised manner how to control their own neural state,
spatially represented in EEG channels, immediately during each
no-delay session. Contrastingly, participants’ subjective agency
was not modulated (i.e., reduced) even by a 20-s–delay, unlike
many behavioral studies, given that they should not have been
aware of the inserted delay. This indicates that we do not have an
internal model of our own neural state (EEG state in our case)
by default. Without this type of externalizing BCI system, it is
difficult for us to estimate our own neural state or to compare the
actual and desired states in forward modeling. This should be an
essential factor why we do not feel a sense of control over BCI
interactions in real time, where only outcome-dependent agency
may be elicited (Evans et al., 2015). However, the current study
further indicates that the use of an NF system that implements
real-time sensory feedback like ours in long-term training for
some months (such as daily motor behaviors), for example, an
internal model over that system with subjective online agency can
be developed. If the NF system represents our neural dynamics in
nature, controlling or internalizing the system would mean that
we could regulate our own neural state through the developed
internal model (possibly even without the NF system). The
EEGms, in this sense, are candidates for representing natural
dynamics in the brain. Therefore, developing a BCI-based NF
system in terms of online agency (or internal model) should be
an important framework for future studies. For that purpose,
our results suggest that minimizing feedback delays could
be a worthwhile implementation for improving task learning
(Belinskaya et al., 2020) such as motor learning (Tanaka et al.,
2011).
In comparison with the preceding EEGms-NF system (Diaz
Hernandez et al., 2016), there are several points to be discussed
here. First, the templates in the system as the desired target
state are group-wise topographies in our case, while in the
previous study they were individually defined. This difference
could be expected to result in different training outcomes. In
our case, the spatial pattern itself was the target for participants’
learning, given that the individual EEGms must be slightly
different from the group-wise EEGms (the “canonical” EEGms).
We may observe the changes for individual EEGms moving
toward group-wise EEGms through long-term training in future
studies. However, individually defined templates should be
easier for participants to induce, because they exhibit these
patterns by default in nature, although the canonical EEGms are
reportedly common across participants. Another difference in
the implementation is sensory feedback. For supervised learning,
participants were visually aware of their current and targeted
state simultaneously. In addition to the visual feedback used
in the previous study, the auditory feedback was provided to
indicate Hits (successes in minimizing a prediction error) and
to reward participants, as well. This audio-visual BCI monitors
and externalizes whole-brain mass neural activity as EEGms to
be regulated (Diaz Hernandez et al., 2016). This implementation
enables us to treat four targets simultaneously (see Figure 1).
In addition, the optimal frequency of NF is an important
factor for participants’ learning so that the definition of EEGms
has been further parameterized in terms of its duration and
topographical robustness. Therefore, the difference to note is
our definition of the state. Our system detects only spatio-
temporally robust EEGms (approximately 100 ms for duration
and rthr = 0.8 for spatial similarity) so that strict online
denoising was not necessary, which is always a nuisance in the
“winner-take-all” definition of EEGms. Both noisy states and
spatio-temporally less-robust states are ignored in our system.
Frontiers in Systems Neuroscience | www.frontiersin.org 11 February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 12
Asai et al. Developing EEGms-NF System
FIGURE 9 | Participants’ relative performance for target specificity. (A) The baseline is the no-target period (“task-ready”) under the no-delay condition. (B) The
baseline is the session average (the column-means for “task-general”) for each delay condition. The graphs show the relative Hits (self-recursive arrows for target
specificity) and Misses (other arrows) where the nodes (A, B, C, and D) indicate the target. (C) The bar plot summarizes (B) with ±1 S.E. error bars.
This largely helped the real-time implementation. However, if
a polarity reversal or rapid transition is observed during the
100 ms (e.g., msA followed by msC and msD), the system
ignores that state (i.e., as a mixture of A, C, and D, see
Figure 3B). This is contrasted to a conventional offline EEG
microstate analysis that has been implemented to handle this
with polarity-ignored labeling (msABCD) and final temporal
smoothing among labels (Poulsen et al., 2018). Recently, this
classical approach of microstate labeling (“winner-take-all”) has
been controversial (e.g., Shaw et al., 2019; Mishra et al., 2020).
Some previous studies have intentionally introduced “unlabeled”
states for this reason (e.g., Zanesco et al., 2021). In this sense,
the current system was not developed to be matched with
the classical labels, and there should be non-small differences
between them depending on the analysis parameters, especially
temporal smoothing of labels. Accordingly, our conservative
system was not designed to detect every robust state but to
not falsely detect them. We can imagine that 1-ms or 10-ms–
averaged state is less reliable for learning and needs further
temporal smoothing during a certain period. In this sense, we
may update to a 100-Hz system using overlapping 100-ms blocks
in a future study.
Regarding the potential differences among the four EEGms
in responsible cognitive functionality, our results suggest some
Frontiers in Systems Neuroscience | www.frontiersin.org 12 February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 13
Asai et al. Developing EEGms-NF System
results congruent with previous studies as well as some new
findings, especially in terms of neural controllability. First, for
msD, the most frequently examined among the four EEGms,
was reportedly increased through NF training for 10 days
(Diaz Hernandez et al., 2016). Although that study was the
first attempt to modify EEGms, msD is a specific target for
future clinical application, since msD may be a biomarker
for schizophrenia. Therefore, a link between schizophrenic
symptomatology and the function of msD has often been
discussed. Previous studies indicate that msD is functionally
related to attentional control or central executive function that
the frontoparietal network underpins. Simultaneous EEG-fMRI
recordings suggest a link where the fluctuation of msD (time-
series of spatial correlation to msD template) explains BOLD
signals in the right-lateralized dorsal and ventral areas of the
frontoparietal cortex (Britz et al., 2010). More recently, a simple
mental arithmetic calculation increased participants’ duration
of msD (Seitzman et al., 2017;Bréchet et al., 2019). If msD
is responsible for this attentive cognitive load in general, our
results are consistent with this. Even during 20-s–delay NF
sessions (participants were not aware of this), participants tried
to make more Hits so that the msD was increased compared
to baseline. However, msC has often been discussed in contrast
to msD as some previous studies indicate that the increased
msD coincides with the decreased msC (e.g., Seitzman et al.,
2017;Bréchet et al., 2019). Indeed, simultaneous recordings
suggested that msC is neurally related to the default mode or
saliency network, which is the functionally opposite network
to the frontoparietal network (Britz et al., 2010; as a review,
Seitzman et al., 2017;Michel and Koenig, 2018). In the current
study, msC was the most delay-sensitive and controllable in real
time. In this sense, saliency-related cognitive function may be a
candidate for msC.
In contrast to the msC-D axis for controllable cognitive
states, msA and B have been discussed as more perceptual
functionalities. Simultaneous recordings suggest that msA is
neurally related to the auditory region, while msB is related
to the visual region (Britz et al., 2010). Recent studies also
indicate msA for auditory perception as well as msB for visual
perception, depending on the stimuli of modality or perceptual
experiences (e.g., Cai et al., 2018;Bréchet et al., 2019;D’Croz-
Baron et al., 2021). On the other hand, our results suggest
that msA or B did not increase, even with real-time NF. These
results indicate that the functionality of msA and B (or msA-
B axis) is more perceptual; therefore, it might be passive and
difficult to control intentionally. Although a more comprehensive
understanding of four canonical EEGms is needed, since previous
findings were given in a relatively sporadic matter, our results
discriminated between msC-D and msA-B in terms of real-
time controllability with NF. Our experiment did not examine
each functionality directly and we do not assume that cognitive
functionality has a 1:1 correspondence to each microstate, rather,
microstate “dynamics” could be responsible to our cognition
and perception (Michel and Koenig, 2018;Ruggeri et al., 2019,
2020). However, the result is interpretable in line with or
extending the existing understanding of EEGms where the
functionality of msA and B are perceptual (auditory for msA
and visual for msB), and therefore, may be less controllable,
while msC and D are cognitive (saliency or default mode
for msC and attentive or executive for msD), and therefore,
more controllable.
This study has some limitations for future applications.
The current study was conducted within a single day for
each participant. Accordingly, the effects of long-term training
using the system are unknown, especially for the target
specificity. Even within a session, the total Hit performance
(and msC and msD) increased above the “task-ready” baseline
(Figure 9A), suggesting that long-term training is predicted.
If msC or msD is a candidate biomarker for some diseases
(for example, schizophrenia), our system is clinically applicable
for future studies as it is (without the inserted delay, of
course). Although msA and msB appear more difficult to
control according to the current results, they are still useful
for examining the effect of long-term training, since a non-
significant tendency for increasing was observed (Figures 7,9).
Another possible future approach would be to manipulate a
parameter of threshold (spatial correlation r) individually, as
well as msABCD, respectively. The former serves to absorb
individual differences in an EEG cap condition, head fitness
with a cap, and innate similarity with canonical templates.
The latter serves to equalize the feedback frequency among
msABCD because msC might occur more often innately.
Examining thresholds could improve the balance between
discriminability and detectability among msA, B, C, and D
(Figures 3A,B). For this purpose, pre-training recording is
required to determine the optimal thresholds and objective
procedure. Whether this parameter should be changed daily
through long-term training should also be considered. Even
in this case, our system has already parameterized the
threshold value. This should prove helpful, especially for specific
participants with disorders who might have a reduced msD,
for example, the system with a regular parameter should
be less responsive to them than to healthy participants.
Finally, the definition of a state as a canonical EEGms
and the number of microstates are controversial (Michel
and Koenig, 2018;Tarailis et al., 2021). Further studies
should address this issue, since the current study used
traditional definitions. Because our system was developed to
make the template easily replaceable, new state definitions
can be readily applied for participant learning (e.g., a 7-
EEGms model).
In conclusion, the aim of the current study was to develop a
closed-loop NF system and validate it as a function of the inserted
delay. If real-time NF is effective for learning, the participants’ Hit
performance should increase above baseline and from delayed
conditions. Our results suggest that this is the case, although the
target specificity is still unclear. Future studies can examine long-
term training with this NF system, even for a specific population.
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be
made available by the authors, without undue reservation.
Frontiers in Systems Neuroscience | www.frontiersin.org 13 February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 14
Asai et al. Developing EEGms-NF System
ETHICS STATEMENT
The studies involving human participants were reviewed and
approved by the Ethics Committee of ATR (reference number
is 20–144). The patients/participants provided their written
informed consent to participate in this study.
AUTHOR CONTRIBUTIONS
TA mainly designed the experiments, analyzed the
data, and wrote the manuscript. TH and SK partially
analyzed the data and wrote the manuscript. TA, TH,
SK, and HI discussed the results and reviewed the
manuscript. All authors approved the final version of the
manuscript for submission.
FUNDING
This study was mainly supported by “Research and development
of technology for enhancing functional recovery of elderly and
disabled people based on non-invasive brain imaging and robotic
assistive devices (16824707)”, the Commissioned Research
of National Institute of Information and Communications
Technology (NICT), JAPAN. HI was also partially supported by
JSPS KAKENHI (grant 19H05725).
ACKNOWLEDGMENTS
We would like to thank Chizuru Homma, Yuki Inoue, Yoshiko
Umemoto, Masaru Nishioka, and Ayako Tsukamoto for their
experimental support.
REFERENCES
Alkoby, O., Abu-Rmileh, A., Shriki, O., and Todder, D. (2018). Can we predict
who will respond to neurofeedback? A review of the inefficacy problem and
existing predictors for successful EEG neurofeedback learning. Neuroscience
378, 155–164. doi: 10.1016/j.neuroscience.2016.12.050
Asai, T. (2016). Agency elicits body-ownership: proprioceptive drift toward a
synchronously acting external proxy. Exp. Brain Res. 234, 1163–1174. doi: 10.
1007/s00221-015-4231-y
Asai, T., and Tanno, Y. (2007). The relationship between the sense of self-agency
and schizotypal personality traits. J. Mot. Behav. 39, 162–168. doi: 10.3200/
JMBR.39.3.162-168
Belinskaya, A., Smetanin, N., Lebedev, M. A., and Ossadtchi, A. (2020). Short-delay
neurofeedback facilitates training of the parietal alpha rhythm. J. Neural Eng.
17:abc8d7. doi: 10.1088/1741-2552/abc8d7
Bréchet, L., Brunet, D., Birot, G., Gruetter, R., Michel, C. M., and Jorge,
J. (2019). Capturing the spatiotemporal dynamics of self-generated, task-
initiated thoughts with EEG and fMRI. Neuroimage 194, 82–92. doi: 10.1016/
j.neuroimage.2019.03.029
Britz, J., Van De Ville, D., and Michel, C. M. (2010). BOLD correlates of EEG
topography reveal rapid resting-state network dynamics. Neuroimage 52, 1162–
1170. doi: 10.1016/j.neuroimage.2010.02.052
Brodbeck, V., Kuhn, A., von Wegner, F., Morzelewski, A., Tagliazucchi, E., Borisov,
S., et al. (2012). EEG microstates of wakefulness and NREM sleep. Neuroimage
62, 2129–2139. doi: 10.1016/j.neuroimage.2012.05.060
Cai, Y., Huang, D., Chen, Y., Yang, H., Wang, C. D., Zhao, F., et al. (2018). Deviant
dynamics of resting state electroencephalogram microstate in patients with
subjective tinnitus. Front. Behav. Neurosci. 12:122. doi: 10.3389/fnbeh.2018.
00122
Caspar, E. A., De Beir, A., Lauwers, G., Cleeremans, A., and Vanderborght, B.
(2021). How using brain-machine interfaces influences the human sense of
agency. PLoS One 16:e0245191. doi: 10.1371/journal.pone.0245191
Changeux, J. P., and Michel, C. M. (2006). “Mechanisms of neural integration
at the brain-scale level: the neuronal workspace and microstate models in
Microcircuits,” in The Interface Between Neurons And Global Brain Function,
eds S. Grillner and A. M. Graybiel (Cambridge, MA: MIT Press), 347–370.
D’Croz-Baron, D. F., Bréchet, L., Baker, M., and Karp, T. (2021). Auditory and
visual tasks influence the temporal dynamics of EEG microstates during post-
encoding rest. Brain Topogr. 34, 19–28. doi: 10.1007/s10548-020-00802-4
da Cruz, J. R., Favrod, O., Roinishvili, M., Chkonia, E., Brand, A., Mohr, C., et al.
(2020). EEG microstates are a candidate endophenotype for schizophrenia. Nat.
Commun. 11:3089. doi: 10.1038/s41467-020- 16914-1
de Bock, R., Mackintosh, A. J., Maier, F., Borgwardt, S., Riecher-Rössler, A.,
and Andreou, C. (2020). EEG microstates as biomarker for psychosis in
ultra-high-risk patients. Transl. Psychiatry 10:300. doi: 10.1038/s41398-020-
00963-7
Delorme, A., and Makeig, S. (2004). EEGLAB: an open source toolbox for analysis
of single-trial EEG dynamics including independent component analysis. J.
Neurosci. Methods 134, 9–21. doi: 10.1016/j.jneumeth.2003.10.009
Diaz Hernandez, L., Rieger, K., Baenninger, A., Brandeis, D., and Koenig, T.
(2016). Towards using microstate-neurofeedback for the treatment of psychotic
symptoms in schizophrenia. A feasibility study in healthy participants. Brain
Topogr. 29, 308–321. doi: 10.1007/s10548-015- 0460-4
Evans, N., Gale, S., Schurger, A., and Blanke, O. (2015). Visual feedback dominates
the sense of agency for brain-machine actions. PLoS One 10:e0130019. doi:
10.1371/journal.pone.0130019
Gallagher, I. (2000). Philosophical conceptions of the self: implications for
cognitive science. Trends Cogn. Sci. 4, 14–21. doi: 10.1016/s1364-6613(99)
01417-5
JASP Team (2019). JASP (Version 0.14)[Computer software].
Khanna, A., Pascual-Leone, A., and Farzan, F. (2014). Reliability of resting-state
microstate features in electroencephalography. PLoS One 9:e114163. doi: 10.
1371/journal.pone.0114163
Khanna, A., Pascual-Leone, A., Michel, C. M., and Farzan, F. (2015). Microstates in
resting-state EEG: current status and future directions. Neurosci. Biobehav. Rev.
49, 105–113. doi: 10.1016/j.neubiorev.2014.12.010
Koenig, T., Prichep, L., Lehmann, D., Sosa, P. V., Braeker, E., Kleinlogel, H., et al.
(2002). Millisecond by millisecond, year by year: normative EEG microstates
and developmental stages. Neuroimage 16, 41–48. doi: 10.1006/nimg.2002.1070
Koukkou, M., and Lehmann, D. (1987). An information-processing perspective of
psychophysiological measurements. J. Psychophysiol. 1, 109–112.
Lehmann, D. (1992). “Brain,” in Electric Fields And Brain Functional States. Springer
Proceedings In Physics. Berlin Heidelberg: Springer, 235–248. doi: 10.1007/978-
3-642-84781-3_12
Lehmann, D., Ozaki, H., and Pal, I. (1987). EEG alpha map series: brain
micro-states by space-oriented adaptive segmentation. EEG Alpha Map Series.
.Electroencephalogr. Clin. Neurophysiol. 67, 271–288. doi: 10.1016/0013-
4694(87)90025-3
Lehmann, D., Strik, W. K., Henggeler, B., Koenig, T., and Koukkou, M.
(1998). Brain electric microstates and momentary conscious mind states
as building blocks of spontaneous thinking: i. Visual imagery and abstract
thoughts. Int. J. Psychophysiol. 29, 1–11. doi: 10.1016/s0167-8760(97)00
098-6
Lubianiker, N., Goldway, N., Fruchtman-Steinbok, T., Paret, C., Keynan, J. N.,
Singer, N., et al. (2019). Process-based framework for precise neuromodulation.
Nat. Hum. Behav. 3, 436–445. doi: 10.1038/s41562-019-0573-y
Michel, C. M., and Koenig, T. (2018). EEG microstates as a tool for studying the
temporal dynamics of whole-brain neuronal networks: a review. Neuroimage
180, 577–593. doi: 10.1016/j.neuroimage.2017.11.062
Milz, P., Pascual-Marqui, R. D., Lehmann, D., and Faber, P. L. (2016). Modalities
of thinking: state and trait effects on cross-frequency functional independent
brain networks. Brain Topogr. 29, 477–490. doi: 10.1007/s10548-016-0469- 3
Frontiers in Systems Neuroscience | www.frontiersin.org 14 February 2022 | Volume 16 | Article 786200
fnsys-16-786200 February 21, 2022 Time: 13:59 # 15
Asai et al. Developing EEGms-NF System
Mishra, A., Englitz, B., and Cohen, M. X. (2020). EEG microstates as a continuous
phenomenon. Neuroimage 208:116454. doi: 10.1016/j.neuroimage.2019.116454
Mulholland, T., Boudrot, R., and Davidson, A. (1979). Feedback delay and
amplitude threshold and control of the occipital EEG. Biofeedback Self Regul.
4, 93–102. doi: 10.1007/BF01007104
Murphy, M., Whitton, A. E., Deccy, S., Ironside, M. L., Rutherford, A., Beltzer,
M., et al. (2020). Abnormalities in electroencephalographic microstates are state
and trait markers of major depressive disorder. Neuropsychopharmacology 45,
2030–2037. doi: 10.1038/s41386-020- 0749-1
Pei, G., Guo, G., Chen, D., Yang, R., Shi, Z., Wang, S., et al. (2020). BrainKilter: a
real-time EEG analysis platform for neurofeedback design and training. IEEE
Access 8, 57661–57673. doi: 10.1109/ACCESS.2020.2967903
Perrottelli, A., Giordano, G. M., Brando, F., Giuliani, L., and Mucci, A. (2021).
EEG-based measures in at-risk mental state and early stages of schizophrenia: a
systematic review. Front. Psychiatry 12:653642. doi: 10.3389/fpsyt.2021.653642
Poulsen, A. T., Pedroni, A., Langer, N., and Hansen, L. K. (2018). Microstate
EEGLAB toolbox: an introductory guide. bioRxiv [Preprint]. doi: 10.1101/
289850
Renard, Y., Lotte, F., Gibert, G., Congedo, M., Maby, E., Delannoy, V., et al.
(2010). OpenViBE: an open-source software platform to design, test and use
brain–computer interfaces in real and virtual environments. Presence 19, 35–53.
doi: 10.1162/pres.19.1.35
Ruggeri, P., Meziane, H. B., Koenig, T., and Brandner, C. (2019). A fine-grained
time course investigation of brain dynamics during conflict monitoring. Sci.
Rep. 9:3667. doi: 10.1038/s41598-019- 40277-3
Ruggeri, P., Nguyen, N., Pegna, A. J., and Brandner, C. (2020). Interindividual
differences in brain dynamics of early visual processes: impact on score accuracy
in the mental rotation task. Psychophysiology 57:e13658. doi: 10.1111/psyp.
13658
Seitzman, B. A., Abell, M., Bartley, S. C., Erickson, M. A., Bolbecker, A. R., and
Hetrick, W. P. (2017). Cognitive manipulation of brain electric microstates.
Neuroimage 146, 533–543. doi: 10.1016/j.neuroimage.2016.10.002
Shaw, S. B., Dhindsa, K., Reilly, J. P., and Becker, S. (2019). Capturing the forest but
missing the trees: microstates inadequate for characterizing shorter-scale EEG
dynamics. Neural Comput. 31, 2177–2211. doi: 10.1162/neco_a_01229
Tanaka, H., Homma, K., and Imamizu, H. (2011). Physical delay but not subjective
delay determines learning rate in prism adaptation. Exp. Brain Res. 208, 257–
268. doi: 10.1007/s00221-010- 2476-z
Tarailis,P., Šimkutë, D., Koenig, T., and Griškova-Bulanova, I. (2021). Relationship
between spatiotemporal dynamics of the brain at rest and self-reported
spontaneous thoughts: an EEG microstate approach. J. Pers. Med. 11:1216.
doi: 10.3390/jpm11111216
Tomescu, M. I., Rihs, T. A., Rochas, V., Hardmeier, M., Britz, J., Allali, G., et al.
(2018). From swing to cane: sex differences of EEG resting-state temporal
patterns during maturation and aging. Dev. Cogn. Neurosci. 31, 58–66. doi:
10.1016/j.dcn.2018.04.011
Wackermann, J., Lehmann, D., Michel, C. M., and Strik, W. K. (1993). Adaptive
segmentation of spontaneous EEG map series into spatially defined microstates.
Int. J. Psychophysiol. 14, 269–283. doi: 10.1016/0167-8760(93)90041-m
Watanabe, T., Sasaki, Y., Shibata, K., and Kawato, M. (2017). Advances in fMRI
real-time neurofeedback. Trends Cogn. Sci. 21, 997–1010. doi: 10.1016/j.tics.
2017.09.010
Yamada, T., Hashimoto, R. I., Yahata, N., Ichikawa, N., Yoshihara, Y., Okamoto,
Y., et al. (2017). Resting-state functional connectivity-based biomarkers and
functional MRI-based neurofeedback for psychiatric disorders: a challenge for
developing theranostic biomarkers. Int. J. Neuropsychopharmacol. 20, 769–781.
doi: 10.1093/ijnp/pyx059
Zanesco, A. P. (2020). EEG electric field topography is stable during moments
of high field strength. Brain Topogr. 33, 450–460. doi: 10.1007/s10548-020-
00780-7
Zanesco, A. P., Skwara, A. C., King, B. G., Powers, C., Wineberg, K., and Saron,
C. D. (2021). Meditation training modulates brain electric microstates and felt
states of awareness. Hum. Brain Mapp. 42, 3228–3252. doi: 10.1002/hbm.25430
Zich, C., Debener, S., Kranczioch, C., Bleichner, M. G., Gutberlet, I., and De Vos, M.
(2015). Real-time EEG feedback during simultaneous EEG–fMRI identifies the
cortical signature of motor imagery. Neuroimage 114, 438–447. doi: 10.1016/j.
neuroimage.2015.04.020
Conflict of Interest: The authors declare that the research was conducted in the
absence of any commercial or financial relationships that could be construed as a
potential conflict of interest.
Publisher’s Note: All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated organizations, or those of
the publisher, the editors and the reviewers. Any product that may be evaluated in
this article, or claim that may be made by its manufacturer, is not guaranteed or
endorsed by the publisher.
Copyright © 2022 Asai, Hamamoto, Kashihara and Imamizu. This is an open-access
article distributed under the terms of the Creative Commons Attribution License
(CC BY). The use, distribution or reproduction in other forums is permitted, provided
the original author(s) and the copyright owner(s) are credited and that the original
publication in this journal is cited, in accordance with accepted academicpractice. No
use, distribution or reproduction is permitted which does not comply with theseterms.
Frontiers in Systems Neuroscience | www.frontiersin.org 15 February 2022 | Volume 16 | Article 786200
