This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

IEEE TRANSACTIONS AND JOURNALS TEMPLATE 1

Trial-by-trial Variability of TMS-EEG in Healthy

Controls and Patients with Depressive Disorder
Zikang Niu, Lina Jia, Yang Li, Lijuan Yang, Yi Liu, Siyuan Lian, Dan Wang, Wen Wang, Liu Yang,
Weigang Pan, Xiaoli Li

Abstract—Objective: Depressive disorder has been exploring the neurophysiological mechanism of DE is critical
known to be associated with high variability in resting-state to help doctors and researchers understanding this disease and
electroencephalography (EEG) signals. However, this optimize treatment therapy. Over the past few decades,
phenomenon is often ignored in stimulus-related brain
noninvasive electroencephalogram (EEG) has been favored by
activities. This study proposed a new method to explore the
EEG variability evoked by transcranial magnetic stimulation doctors and researchers for its high accuracy, safety, simplicity,
(TMS, TMS-EEG) in depressive disorder (DE) patients. and high temporal resolution in assessing DE [3-5].
Methods: The TMS-EEG data were collected from 34 DE Previous studies have indicated that the dorsolateral
patients and 36 healthy controls (HC). The maximum prefrontal cortex (DLPFC) plays a significant role in the
eigenvalue of the real binary correlation matrix, calculated pathophysiology of DE [6]. Lots of studies have demonstrated
between different trials using cross-correlation and
that synchronized EEG data, evoked by Transcranial Magnetic
surrogate methods, was extracted to assess trial-by-trial
variability (TTV) of TMS-EEG. Results: The new method was Stimulation (TMS, TMS-EEG) on the DLPFC, can reflect the
found to more sensitive and reliable than the standard cortical excitability, inhibition, plasticity, and the changes of
deviation method. DE patients exhibited significantly smaller connectivity patterns in DE patients [7-9]. It has been observed
TTV in Gamma band and greater TTV in Delta band than HC. that the amplitude of N100 and N45 components of DE patients
Furthermore, the HAMD-17 scores were negatively is lower than normal individuals, and these two components
correlated with TTV values in Gamma band. Conclusions:
exhibit high accuracy in predicting disease state [10, 11]. In
This study represented the first investigation into the TTV in
TMS-EEG data and revealed abnormal values in DE patients. addition to being biomarkers for disease diagnosis, TMS-EEG
Those findings enhance our understanding of TMS-EEG excitability indicators can also predict the treatment effect of
technology and provide valuable insights for studying the depression and monitor its biological response. For instance,
characteristics of DE. after a Magnetic Seizure Therapy, a decrease of N100 was
observed in patients with refractory depression, and this change
Index Terms— TMS-EEG, Trial-by-trial Variability, associated with the Scale for Suicide Ideation. The accuracy,
Depressive Disorder.
sensitivity, and specificity of this index in predicting suicidal
I. INTRODUCTION ideation reached 89%, 90%, and 89%, respectively [12]. In
addition to being excitatory indicators, TMS-induced brain

A
s a common and recurrent disorder, depressive
disorder (DE) is typically characterized by a lowered connection activity can also be utilized as a biomarker for
mood, reduced energy, and lowered enjoyment. Even detecting and diagnosing depression. Electroconvulsive
Therapy studies have found that phase locking value induced
with adequate full-course medication, most adult patients
cannot effectively achieve remission[1, 2]. In the clinical by single TMS pulse in EEG can characterize depressive states
diagnosis and evaluation of intervention effects, scales such as and evaluate the antidepressant effects of neuromodulation
HAMD-17 are deemed inefficient and lack objectivity due to [13].
their reliance on the doctors’ mature experience. Therefore, Neural variability is a robust phenomenon which has been

This work was supported in part by the STI2030 Major Projects (No. Lina Jia is with Beijing Key Laboratory of Mental Disorders, Beijing
2021ZD0204300); in part by the Scientific Research Initiated by Beijing Anding Hospital, Capital Medical University, Beijing, 100088, China and
Anding Hospital Researchers (No. 8-202173FS-2); and in part by the Xuanwu Hospital Capital Medical University, Beijing, 100053, China (e-
Beijing Municipal Hospital Research and Cultivation Project (No. mail: [email protected]).
PX2024069). (Zikang Niu and Lina Jia contributed equally to this work). Yang Li, Lijuan Yang, Yi Liu, Siyuan Lian, Dan Wang, Wen Wang,
(Corresponding authors: Liu Yang, Weigang Pan; Xiaoli Li, Weigang and Weigang Pan are with Beijing Key Laboratory of Mental Disorders,
Pan is the lead contact). Beijing Anding Hospital, Capital Medical University, Beijing, 100088,
This work involved human subjects or animals in its research. China (e-mail: [email protected]; 1058669692@ qq.com;
Approval of all ethical and experimental procedures and protocols was [email protected]; [email protected]; wangdandoc@cc
granted by the Beijing Anding Hospital (NO. 202173FS-2) and Beijing mu.edu.cn; [email protected]; [email protected] ).
Normal University (NO. CNL_A_0010_005). All subjects have signed Liu Yang is with Aviation Psychology Research Office, Air Force
the informed consents. Medical Center, Beijing, 100142, China (e-mail: [email protected]
Zikang Niu is with Aviation Psychology Research Office, Air Force om).
Medical Center, Beijing, 100142, China and State Key Laboratory of Xiaoli Li is with Guangdong Artificial Intelligence and Digital Economy
Cognitive Neuroscience and Learning, Beijing Normal University, Laboratory (Guangzhou), Guangzhou 510335, China and School of
Beijing, 100875, China (e-mail: [email protected]). Automation Science and Engineering, South China University of
Technology, Guangzhou, 510641, China. (e-mail: [email protected] ).

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

observed in intracellular membrane potential, extracellular TABLE I

DEMOGRAPHIC AND CLINICAL DATA
recordings of spiking activity, and human EEG,
electrocorticography, magnetoencephalography, and functional Group HC DE
magnetic resonance imaging studies [14-16]. In resting-state Sample size 36 34
EEG studies, nonlinear correlation analysis showed that Age(mean±std) 33.72z±12.21 33.44±11.67
depressive symptoms were associated with high variability of Sex(male/female) 5/31 3/31
resting-state EEG signals [17, 18]. For instance, Jaworska et al
HAMD-17 25.09±5.70
found reduced multiscale entropy at fine temporal scales,
HC: healthy control; DE: depressive disorder, HAMD-17: Hamilton
especially in frontal-central region, and the increased multiscale Depression Scale -17 items.
entropy value diffusely distributed in coarser temporal scales,
which was related to the magnitude of the antidepressant TTV of TMS-EEG.
treatment response [19]. Lee and his colleagues found that the
severity of major depressive disorder had a positive correlation II. MATERIAL AND METHODS
with the long-range temporal autocorrelation of resting-state A. Participants
EEG [20]. Indexes calculated based on Higuchi's fractal We recruited 34 DE patients and 36 age-matched HC in
dimension and Lempel-Ziv complexity correlation analysis Beijing Anding Hospital and Beijing Normal University. All
indicated that the complexity of brain activity in DE patients participants were right hand, aged between 18 and 60 years old,
was greater than that of healthy controls (HC), indicating a and Wilcoxon’s signed test results showed no significant age
greater variability [21]. In recent years, studies have found that differences between groups (detailed information in Table I).
Fuzzy Measure Entropy, an improved method based on Fuzzy The DE patients were diagnosed according to ICD-10 criteria
Entropy, was related to the severity of the disease [22]. and didn’t have any other mental illnesses. In addition,
In event-related potential studies, researchers typically individuals with a history of neurological disorder, seizures,
analyze and discussed the data by averaging over trials. This head injury and other conditions that made them unsuitable for
averaging operation requires the assumption of linear receiving TMS stimulation were excluded. The Hamilton
superposition between basic random ongoing background Depression Scale-17 item (HAMD-17) of DE patients were
activity and highly stereotyped, repeatable evoked responses used to assess the severity of depression in DE patients by two
[23, 24]. However, numerous studies have demonstrated that trained doctors who underwent consistency training for
the neural response to repeated identical stimulation is highly assessments. Finally, patients with a HAMD-17 score greater
variable [25-27]. One possible approach to measure neural than 17 were enrolled in the study and TMS-EEG data were
variability is to calculate the trial-by-trial variability (TTV), as collected within 24 hours after the HAMD-17 evaluation. Due
demonstrated in animal and human studies [16, 28]. Several to the actual situation of patients and hospital, only medicated
pieces of evidence suggest that TTV can impact behavioral patients were included in the study without controlling the type
performance. First, greater TTV quenching in sensory cortices of medicine. To minimize the potential influence of medicine
was associated with better perceptual performance [29]. on the results, only patients who had taken the medicine steadily
Second, the behavioral performance can be improved by for more than half a month were recruited and their TMS-EEG
reducing the trial-by-trial response variability of the visual data were collected at least 8 hours after taking their medicine.
cortex in visual-related attention experiments[30]. In the DE,
those activities are usually assessed through visual and auditory B. Experimental equipment and procedure
pathways. For instance, the inter-trial phase coherence of 40 Based on a large number of previous studies[33-35], we
Hz-auditory steady-state responses may serve as potential utilized the classical experimental equipment and single pulse
neurophysiological markers for early depression detection, and stimulation paradigm, which have been previously employed in
aid in understanding the underlying symptom severity in first- the HC and DE patients. In this experiment, we used a Rapid2
episode major depressive disorder [31]. Combining the steady- system with D70 coil (Magstim, UK) to stimulate the DLPFC
state topographical probe and 13Hz steady-state visually of participants’ brain. Additionally, 64-channel
evoked potentials, Kemp et al. discovered that during acute synchronization EEG data (EASYCAP GmbH and BrainAmp
serotonergic augmentation in a serotonergic antidepressant MR amplifier, Brian Products, Germany) were collected in an
research, the response to pleasant valences of DE was electrically and acoustically shielded room at a sampling rate of
potentiated while the response to unpleasant valences was 2500 Hz.
suppressed [32]. During the experiment, all electrode impedances were
However, the traditional TMS-EEG analysis, which was maintained below 5kΩ and participants were instructed to sit
based on averaging all EEG trials, have ignored the neural comfortably in a chair with their arms at rest. First, located C3
variability between repeatable stimulations. In order to position by 10-20 system, and determined the M1, which can
investigate the TTV of TMS-EEG data and the abnormal be induced the max motor‐evoked potentials in the first dorsal
variability in DE patients, we collected TMS-EEG data of DE interosseous muscle, with a 45°angle between the coil handle
patients and HC. Based on the cross correlation and surrogate and Anterior posterior sagittal line, by using the conventional
method, we extracted the maximum eigenvalue of the real nonnavigated strategies in Roland’s paper[36]. Then
binary correlation matrix among different trial data to assess the stimulated the M1 to determine the resting motor threshold,

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

which was defined as the minimum stimulation intensity to 1

evoke 50μV peak-to-peak motor evoked potential amplitude at the unit is ms  0.00195ms ;
512
least 5 of 10 trials in the relaxed first dorsal interosseous muscle After that, computed the discrete stationary wavelet
[37]. After that, stimulate the DLPFC in a similar way to M1, transform using the method employed by Dutt[45]. Then, we
which was located by the BeamF3 method [38], for a total 120
extracted the 1~8 levels’ detail coefficients (CD, CD1~8) in -
times with a 2~2.5s by using the 110% RMT. In order to avoid 200~400ms for further analysis.
possible auditory responses and the bone-conducted sound Since the binary based filtering process of discrete stationary
caused by the TMS click, the participants wore earphone with wavelet transform, Nyquist Sampling Theorem and 512Hz
a white noise masking sound while a thin sponge was placed sampling rate of data, the CD1~8 can reflect the signal X’s
under the coil [39]. information in the 128~256Hz (CD1), 64~128Hz (CD1),
C. TMS-EEG data preprocessing 32~64Hz (CD2), 16~32Hz (CD3), 8~ 16Hz (CD4), 4~8Hz
All EEG data were preprocessed by using the TESA toolbox (CD5), 2~4Hz(CD6) and 1~2Hz(CD7) frequency bands,
[40-42] with the following procedures: (1) Found bad channels respectively.
and replaced them with the superfast spherical interpolation F. Trial-by-trial variability analysis
method; (2) Extracted epochs from -900 to 900ms and
A. Based on the maximum eigenvalue of correlation matrix,
performed baseline correction by using data from −500ms to
we calculated TTV values of the CD1~8. Taking the CD3 series
0ms (3) Removed -2 to 10ms data, which contains TMS pulse
( d ) as an example, the algorithm is as follows:
artifacts, and interpolated them by using the cubic interpolation
method; (4) Rejected bad trials and down sampled data to 512 (1) Extracted the TMS-EEG data in Tpre and Tpost form N
Hz; (5) Performed the first ICA decomposition and removed TME-EEG trials and normalized them using z score method;
muscle, electrical and movement noise components where the Tpre is -200~0ms, Tpost is 0~400ms and N is the
(number=9.765 ± 2.254) from the first 15 components; (6) number of TMS-EEG trial.
Filtered the data by the fourth-order Butterworth band-stop
(2) Calculated the cross correlation Ci , j between two trials
filter (48-52 Hz) and fourth-order Butterworth band-pass filter
(1-100 Hz); (7) Performed the first ICA decomposition and in Tpre period :
removed muscle, blink, eye movement and electrode noise 1 t2

artifact (number=17.838 ± 4.730) from all Ci , j 

M
d
t  t1
i ,t  d j ,t (2)
components(number=43.221±3.689); (10) Converted the data
into average reference. and extracted the data from -250 to where d i ,t and d j ,t is CD3 series of trial i or j ( 1  i, j  N )
500ms, for the further analysis. in time point t ( 1  t  M ), M is the data length of d i ,t from t1
D. TMS-evoked potential analysis to t2, t1  200ms，t2  0ms .
An average TEP signal was obtained from each electrode. (3) In order to test the reliability of cross correlation between
The butterfly outputs of HC and DE are shown in Fig. 1. To trials, generated 50 fake TMS-EEG data by randomly shuffling
further analysis the differences of each component between the each single trial TMS-EEG data in the time ranging from -200
two groups, we also performed 2D topographic scalp mapping to 0ms. And, calculated the cross correlation CFi , j from those
and source estimation analysis.
fake data. Then, transformed the cross correlation matrix into a
In addition, the source location was calculated using
Brainstorm MATLAB toolbox with the follows [43]: (1) Due binary matrix ( CBi , j ):
to the absence of MRI data, co-registered the EEG cap to the 1 if Ci , j  A1i , j
Colin27 head model; (2) Second, used the OpenMEEG to 
CBi , j  0 if A1i , j <Ci , j  A2i , j (3)
computed the geometric head model [44]; (3) Third, calculated 1 if C  A2
the noise covariance matrix based on the -200~-20ms data; (4)  i, j i, j

Finally, constrained the dipoles o the cortex surface and where A1i , j and A2i , j is the 5% and 95% largest value of the
computed the inverse solution by using the current density
analysis. 50 fake TMS-EEG data’s CFi , j between trial i and j;
(4) Decomposed the binary correlation matrix CB using the
E. Discrete stationary wavelet transform analysis eigenvalue decomposition method:
First, since the length of input data must be divided by 2^Q CBvi  i vi (4)
in discrete stationary wavelet transform, we extended the TMS-
EEG data to a length of 512 by using the periodic extension where i is the eigenvalue and  i is the corresponding
method: eigenvector;
S (t )  [ X (64), , X (1), X (1), , X ( H ), (5) Normalized the max eigenvalue METpre by the number of
(1)
X ( H ), , X ( H  64  1)], H  384 trials.
Where X is the original signal of TMS-EEG, H is the length max(i )
METpre  (5)
of X and t is the time variable of TMS-EEG’s trial signal X with N

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

Since CBi , j is a real symmetric matrix with all elements are TABLE Ⅱ
THE DETAIL STATISTICAL VALUES OF TEP
0 or 1, the METpre value are in the range of [1/60, 1] which can
Channel HC(μV) DE(μV) p Value T Value
provide information about the synchronization of neural
P3 -0.645±1.003 -0.164±0.889 0.020 2.092
activity over trials. If the TMS-EEG data series are completely
P5 -1.086±1.149 -0.581±0.900 0.024 2.014
uncorrelated over trials, then CBi , j is an identity matrix. So the
C5 -0.444±0.902 0.166±0.819 0.002 2.918
max(i ) is 1 and MET is equal to 1/60. If the TMS-EEG data
-0.875±0.967 -0.151±0.768
pre
CP5 0.001 3.415
series are completely correlated over trials, then all the elements
T7 -0.916±1.161 -0.162±0.990 0.003 2.883
of CBi , j are equal to 1. So the max(i ) is N and MET is equal
-0.569±1.053 0.008±0.869
pre
FT7 0.008 2.467
to 1.
(5) Repeat steps (2)~(5), extracted the MET of the TMS- post
examine the changes in TTV induced by TMS, we performed
EEG signals in Tpost . cluster based permutation test on VarSTD of TTVSTD before and
(6) Calculated the TTVME : after TMS pulse, as well as on the real correlated trial number
METpost  METpre of TTVME . In the TTVME method, N are set as 10~75 to explore
TTVME  (6)
METpre the influence of trial number on TTVME . Thirdly, we executed
cluster based permutation test to examine the significant
B. To demonstrate the sensitivity of the TTVME , we also
differences of TTVME or TTVSTD value between two
calculated each participant's TMS evoked neural variability
groups[46]. In order to reduce the probability of false positive,
( TTVSTD ) using standard deviation method, which is usually
Bonferroni procedures was performed between two TTV
used in EEG signal variability studies but ignores the similarity calculating method, that is to say, the alpha levels (p-values)
in time dynamic characteristic between different trial series, was set to 0.05/2 = 0.025. Finally, in the correlation analysis
The steps were as follows:
between TTVME or TTVSTD and HAMD-17 of DE, we also
(1) Calculated the mean value of variability in a time period
performed cluster based permutation correlation analysis,
Tpre :
which was similar with the cluster based permutation test,
N between the symptom scores and every electrode.
t2  (d i ,t  dt )

t  t1
i 1
N
(7) Ⅲ. RESULTS
VarSTD T  A. TMS evoked potentials
pre
M
Fig. 1 A and B showed the average TEP and source location
where d t is the mean value of the sequence d i ,t over trials,
analysis of the two groups. After TMS pulse stimulation, all
N is the number of TMS-EEG trials. M is the data length of d i ,t subjects exhibited seven components (P30: 33ms, N45: 44ms,
P60: 73ms, N100: 113ms, P200: 171 or 200ms, N280: 276ms,
from t1 to t2, t1  200ms，t2  0ms .
P380: 389ms). In group level, the latency of DE in the P200
(2) Same as step (1), extracted the VarSTD Tpost of the TMS- component was shorter than that of HC in group level. So the
P200 latency of DE were respectively set as 171ms while it was
EEG signals in Tpost .
set at 200ms for HC. Statistical results indicated that the P200
(3) Calculated the TTVSTD : amplitudes of channel P3, P5, C5, CP5, T7 and FT7 in DE were
VarSTDT  VarSTDT significantly larger than those in HC. The detail values were
TTVSTD  post pre
(8) showed in Table Ⅱ.
VarSTDT
pre
B. Trial-by-trial variability
It should be noticed that the TTVSTD describes
In TTVSTD method, statistical analysis showed that there was
desynchronization variability over TMS-EEG trials while the
TTVME represent the synchronization variability. To describe no significant difference in VarSTD before and after TMS pulse.
the variability in a uniform manner, we used “variability” to For instance, Fig.2A showed the CD3’s VarSTD value in HC.
represent desynchronization variability. In other words, when Compared with HC, DE had lager TTVSTD in the left frontal and
the TTVME is larger, the TTV of TMS-EEG is small. right central-parietal lobe and smaller TTVSTD in the left
G. Statistical analysis central, left parietal and middle frontal lobe. However, no
significant difference was found between DE and HC (Fig.2D).
To study the differences between the TMS-EEG data at HC
and DE, we first conducted cluster based permutation test Fig. 3 showed the TTVME results of two groups. First, to
between two groups on all channels’ TEPs. Subsequently, to investigate the impact of trial number N on TTVME , we
randomly selected 10 to 75 TMS-EEG trials to calculated the

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

Fig. 1. TMS evoked potentials (TEPs) of healthy controls (A) and depression patients (B). Left: Butterfly plot of all electrode waveforms; right:
voltage distribution in 2D and current density map at the cortex in each peak. ”×” indicates p<0.05 between HC and DE in one channel

TABLE Ⅲ
THE DETAIL STATISTICAL VALUES OF TTVME
Channel HC(μV) DE(μV) T Value T Value
T8 0.143±0.246 0.014±0.156 0.021 -2.063
Pz 0.128±0.194 0.033±0.145 0.010 -2.398
CP2 0.126±0.164 0.051±0.136 0.020 -2.103
CP6 0.146±0.299 0.054±0.146 0.024 -2.015
CD3

CP4 0.140±0.234 0.088±0.168 0.023 -2.028

C6 0.132±0.247 0.052±0.137 0.017 -2.155
FT8 0.158±0.223 0.059±0.167 0.011 -2.345
TP8 0.152±0.289 0.045±0.126 0.018 -2.127
CPz 0.148±0.204 0.057±0.155 0.014 -2.252
Pz -0.179±0.166 -0.065±0.200 0.005 2.616
Fig. 2. TTVSTD analysis of CD3. (A) analysis of VarSTD before and
CD7

CP1 -0.158±0.140 -0.073±0.178 0.013 2.261

after TMS pulse in HC; (B) Topographic distribution and statistical of
analysis of TTVSTD between HC and DE in CD3. ”×” indicates p<0.025 CPz -0.161±0.171 -0.082±0.143 0.018 2.129

TTVME and observed that the fluctuation of TTVME remained Compared with HC, DE showed significant smaller values in
stable when N was greater than 30 (Fig. 3 A). In order to capture central-parietal and right temporal lobe. In CD7, HC exhibited
more brain signals, N was set to be 60 for calculating of cross higher value in the right frontal and left temporal lobes while
correlation. We also analyzed the real correlated trial number DE showed higher values in the right parietal and parietal-
of two groups and found that TMS increased the correlation occipital lobe. Compared with HC, DE had a significant larger
over trials. For instance, in HC group, TMS significantly value in the parietal lobe. The detail values were showed in
increased this value of CD3 in lots of channels and the channel Table Ⅲ.
in left central and middle frontal lobe showed the most changes C. Correlation between TTV and HAMD-17
(Fig. 3 B).
The correlation analysis showed there were significant
Statistical analysis revealed significant differences in TTVME
correlations between TTVME and depressive symptoms of DE
of CD3 and CD7 between two groups (Fig. 3 C and D). In CD3,
(Fig. 4). In CD3, the TTVME values of some channels were
HC had a greater TTVME in left frontal and central lobe while
negatively correlated with HAMD-17 scale. Those channels
those values of DE had higher values in left central lobe. were distributed in two clusters: cluster 1 (left parietal region),

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

TTVME analysis. (A) The TTVME value of CD3 for different trial number in channel Cz. (B) analysis of real correlated trial number over
Fig. 3.
TMS-EEG trials before and after TMS pulse in HC. (C) and (D) Topographic distribution and statistical analysis of CD3 and CD7’s TTVME . ”×”
indicates p<0.025.

Fig. 4. Correlation analysis between TTVME of CD3 and clinical HAMD-17. (A) Cluster 1: right central region; (B) Cluster 2: left parietal region.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

TABLE Ⅳ
intensity (110% RMT) may be so large that it failed to evoked
the a significant different early TEP between the DE and HC.
THE DETAIL STATISTICAL VALUES OF TTVME
Channel RHO p Value F Value B. TTVSTD VS TTVME
C4 -0.452 0.007 -2.868 Neural activity is variable over time and across stimulation
C6 -0.439 0.009 -2.762 trials. Among the trials with identical stimulation, neural
Cluster 1 variability is small before stimulus onset and significantly
FC4 -0.464 0.006 -2.959
increased after stimulus presentation. In traditional event-
FC6 -0.419 0.014 -2.612
related potential experiments, the standard deviation method is
P3 -0.392 0.022 -2.409 usually used to describe neural variability of cognitive
P5 -0.356 0.039 -2.154 processing pathways after the brain receives external stimuli
Cluster 2
P7 -0.386 0.024 -2.366 through the sensory organ [53]. In this study, we first studied
CP5 -0.396 0.020 -2.442
the TTV in the TMS-EEG data and developed a new method to
assess the neural variability of TMS-EEG data. The results
showed that TTVME were changed after TMS stimulation. From
C4, C6, FC6 and FC4; cluster 2 (left parietal region), P3, P5, P7
and CP5. The detail values were as shown in TABLE Ⅳ. this phenomenon, we assume that the reduced reproducibility
of neural activity may be attributed to the TMS stimulation on
IV. DISCUSSION DLFPC leading to the depolarization of neurons and evoked
action potentials, which phenomenon can also be observed in
A. TMS evoked potentials
other event-related potential studies and EEG calculation
TEPs, which are evoked by TMS on DLPFC, are typically methods[9, 54, 55]. Our results indicate that TTVME is more
defined as P30, N45, P60, N100, and P200 [47]. Our study also
sensitive for representing abnormal in DE patients and reducing
found these components using supraliminal stimulation patterns
of TMS pulses. The generation of TEPs is associated with the falsity of the TTV index than TTVSTD .
spatial and temporal summation of postsynaptic potentials C. Trial-by-trial variability and Depressive disorder
which originate from a large number of pyramidal neurons and
In the pathophysiology of clinical disorders, oscillations can
interneurons. These components are thought to be important
provide important information about the effects of plasticity-
indices of cortical excitability. For instance, early TEPs (such
inducing protocols on brain activity. For instance, specific
as N45 and P60) are associated with postsynaptic inhibitory
frequency bands can be associated with specific behaviors or
Gamma aminobutyric acid A receptors and glutamatergic
cognition [56]. Gamma band activities are commonly used to
activity, while the N100 component is involved in cortical
characterize depressive disorder and schizophrenia patients
inhibition mediated by presynaptic and postsynaptic Gamma
[57-59]. Event-related potential and resting-state EEG studies
aminobutyric acid B receptors [48, 49].
have shown that Gamma oscillations have a good diagnostic
Previous studies have demonstrated that the TMS-EEG
capability for DE [3, 60]. Unlike traditional event-related
excitability can not only be used as a diagnostic biomarker for
potential, TMS–EEG offers an opportunity to measure those
depression but also to monitor biological responses to
activities in different frequency bands of the brain without
treatment. For instance, compared with HC, major depressive
involving the processing of sensory pathways. Our results
disorder patients exhibit lower N100 and N45 amplitudes and
revealed that DE patients had significantly smaller variability
greater P200 amplitudes [10, 50]. Additionally, Theta burst
of CD3, which mainly reflects the gamma band information
stimulation can reduce the N45 in the right intraparietal lobules,
(32~64Hz), than HC. Pizzagalli and Isomura both found that
while the administration of Baclofen (a GABAB agonist), can
Gamma activities, which was evoked by auditory steady-state
increase the N100 amplitude in healthy participants [48, 51].
responses or Flanker task , can be used to distinguish depressed
Unfortunately, our study only found significant greater P200
patients from healthy controls, even unipolar depression and
amplitude of DE in channel P3, P5, C5, CP5, T7 and FT7, while
no significant difference was observed in N45 or N100. bipolar disorder in their own experiments[54, 61]. In clinical
Previous studies, which set the intensity of TMS pulse to less studies, Gamma oscillations can not only be used to distinguish
than 100%RMT, have reported significant differences in the the depressive disorder, but also can be the potential biomarker
early TEP between DE and HC. However, such differences are for the treatment. For instance, Pellicciari and his colleague
rarely reported when setting the intensity of TMS pulse to larger found the gamma oscillation, which were evoked by TMS in
than 100%RMT [50, 52]. Only Dhami found a significant P30 DE patients, was asymmetrical in bilateral DLPFC, and a 10-
difference among youth with depression[51]. Therefore, we day Theta Burst Stimulation can reduce this abnormal
hypothesis that there are two possible reason. One possible phenomena [62]. In a paired associative stimulation studies,
reason may be the different clinical condition. For example, Noda found that compared to HC, the changes of Gamma power
other studies usually recruited young, fist-episode or severely and Theta-Gamma coupling activities were significantly lower
suicidal patients which this paper are not identical with. In in DE [64]. Our results provide another important evidence for
addition to the clinical factors of patients, the TMS pulse the abnormal Gamma brain activity in depressed patients.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

D. Correlation between TTV and HAMD-17 REFERENCES

In this study, we discovered a significant negative correlation [1] R. S. McIntyre et al., "Treatment-resistant depression: Definitions, review
of the evidence, and algorithmic approach," (in English), Journal of
between the HAMD-17 and TTVME of CD3 which mainly Affective Disorders, vol. 156, no. 1, pp. 1-7, Mar 2014.
reflecting the Gamma(32~64Hz) band information. Similar [2] World Health Organization (WHO), The ICD-10 classification of mental
results have been found in previous studies of patients with DE. and behavioural disorders. World Health Organization, 1993.
[3] F. S. de Aguiar Neto and J. L. G. Rosa, "Depression biomarkers using non-
For instance, one study reported a 50% reduction of Gamma invasive EEG: A review," Neuroscience & Biobehavioral Reviews, vol.
aminobutyric acid in the occipital cortex of major depressive 105, pp. 83-93, Oct 2019.
disorder patients while it is also significantly associated with [4] S. Yasin, S. A. Hussain, S. Aslan, I. Raza, M. Muzammel, and A. Othmani,
"EEG based Major Depressive disorder and Bipolar disorder detection
the severity of depression [63]. High-frequency activity is using Neural Networks:A review," (in English), Computer Methods and
known to be associated with the continuation of the cognitive Programs in Biomedicine, vol. 202, p. 106007, Apr 2021.
setting and the dominance of endogenous top-down influences, [5] H. Zhang et al., "The applied principles of EEG analysis methods in
neuroscience and clinical neurology," Military Medical Research, vol. 10,
which override the effect of potentially novel, or unexpected, no. 1, p. 67, 2023/12/19 2023.
external events [64]. Our results indicated TTVME had a [6] C. R. Chin Fatt et al., "Dorsolateral Prefrontal Cortex and Subcallosal
Cingulate Connectivity Show Preferential Antidepressant Response in
negative correlation with HAMD-17. We hypothesis that our Major Depressive Disorder," Biological Psychiatry: Cognitive
findings may be related to a defect in receiving and processing Neuroscience and Neuroimaging, vol. 6, no. 1, pp. 20-28, Jan 2021.
new information in depressed patients. Future experiments need [7] P. Canali, "A role for TMS/EEG in neuropsychiatric disorders," Neurology,
Psychiatry and Brain Research, vol. 20, no. 2, pp. 37-40, 2014.
to be designed to investigate the difference between TMS-EEG [8] P. Canali et al., "Shared reduction of oscillatory natural frequencies in
and other states. bipolar disorder, major depressive disorder and schizophrenia," Journal of
Affective Disorders, vol. 184, no. 15, pp. 111-115, Sep 15 2015.
V. CONCLUSIONS [9] Z. H. Ding, L. Z. Guan, W. B. He, H. Gu, Y. Wang, and X. L. Li, "Spatial
characteristics of closed-loop TMS-EEG with occipital alpha-phase
In this study, we not only proposed a new, more sensitive, and synchronized," (in English), Biomedical Signal Processing and Control,
reliable method to assess TTV based on the maximum vol. 83, p. 104650, May 2023.
eigenvalue of the cross-correlation matrix, but also conducted [10] D. Voineskos et al., "Altered Transcranial Magnetic Stimulation-
Electroencephalographic Markers of Inhibition and Excitation in the
the first study of TTV in TMS-EEG data. Additionally, we Dorsolateral Prefrontal Cortex in Major Depressive Disorder," (in English),
observed significantly smaller TTV in Gamma band and greater Biological Psychiatry, vol. 85, no. 6, pp. 477-486, Mar 15 2019.
TTV in Delta band in DE patients. Furthermore, the HAMD-17 [11] P. Belardinelli et al., "TMS-EEG signatures of glutamatergic
neurotransmission in human cortex," (in English), Scientific Reports, vol.
scores were negatively correlated with TTV values in Gamma 11, no. 1, p. 8159, Apr 14 2021.
band. Our findings may contribute to a better understanding of [12] Y. M. Sun et al., "Indicators for Remission of Suicidal Ideation Following
the capabilities of TMS-EEG technology and the specific Magnetic Seizure Therapy in Patients With Treatment-Resistant
Depression," (in English), Jama Psychiatry, vol. 73, no. 4, pp. 337-345,
characteristics associated with depressive disorders for Apr 2016.
researchers and medical professionals. However, our study still [13] E. Miyauchi, M. Ide, H. Tachikawa, K. Nemoto, T. Arai, and M. Kawasaki,
has some limitations: 1. Lack of MRI data to help localize the "A novel approach for assessing neuromodulation using phase-locked
information measured with TMS-EEG," (in English), Scientific Reports,
M1 motor region and DLPFC; 2. Lack of a paired-pulse TMS– vol. 9, no. 1, p. 428, Jan 23 2019.
EEG experiment (e.g. long interval intracortical inhibition) to [14] M. M. Churchland et al., "Stimulus onset quenches neural variability: a
examine the inhibition of neural variability; 3. No intensity widespread cortical phenomenon," (in English), Nature Neuroscience, vol.
lower than 100% RMT were used as the TMS pulse intensity. 13, no. 3, pp. 369-U25, Mar 2010.
[15] B. J. He and J. M. Zempel, "Average Is Optimal: An Inverted-U
In future studies, we will recruit more DE patients to validate Relationship between Trial-to-Trial Brain Activity and Behavioral
the findings and apply this method to other patients with Performance," (in English), Plos Computational Biology, vol. 9, no. 11, p.
different brain diseases. Moreover, new studies should be e1003348, Nov 2013.
[16] A. Arazi, G. Gonen-Yaacovi, and I. Dinstein, "The Magnitude of Trial-
designed to resolve these limitations. By-Trial Neural Variability Is Reproducible over Time and across Tasks
in Humans," (in English), Eneuro, vol. 4, no. 6, pp. e0292-17.2017, Nov-
ACKNOWLEDGMENT Dec 2017.
[17] S. Yun and B. Jeong, "Aberrant EEG signal variability at a specific
The authors are particularly thankful to the Beijing Anding temporal scale in major depressive disorder," (in English), Clinical
Hospital and Beijing Normal University for their cooperation in Neurophysiology, vol. 132, no. 8, pp. 1866-1877, Aug 2021.
recruiting the HC and DE patients. They would also like to [18] Z. K. Niu et al., "Scale-free dynamics of microstate sequence in negative
schizophrenia and depressive disorder," (in English), Computers in
thank Xin Liu for their valuable helps on grammar of Biology and Medicine, vol. 143, p. 105287, Apr 2022.
manuscript. [19] N. Jaworska, H. Y. Wang, D. M. Smith, P. Blier, V. Knott, and A. B.
Protzner, "Pre-treatment EEG signal variability is associated with
DECLARATION OF COMPETING INTEREST treatment success in depression," (in English), Neuroimage-Clinical, vol.
17, pp. 368-377, Oct 31 2018.
The authors declare that they have no known competing [20] J. S. Lee, B. H. Yang, J. H. Lee, J. H. Choi, I. G. Choi, and S. B. Kim,
financial interests or personal relationships that could have "Detrended fluctuation analysis of resting EEG in depressed outpatients
appeared to influence the work reported in this paper. and healthy controls," (in English), Clinical Neurophysiology, vol. 118, no.
11, pp. 2489-2496, Nov 2007.
[21] M. Bachmann, J. Lass, A. Suhhova, and H. Hinrikus, "Spectral
Asymmetry and Higuchi's Fractal Dimension Measures of Depression
Electroencephalogram," (in English), Computational and Mathematical
Methods in Medicine, vol. 2013, p. 251638, 2013/10/22 2013.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

[22] M. Kalev K Fau - Bachmann, L. Bachmann M Fau - Orgo, J. Orgo L Fau [42] Y. Wang et al., "Comparison of synchrosqueezing transform to alternative
- Lass, H. Lass J Fau - Hinrikus, and H. Hinrikus, "Lempel-Ziv and methods for time-frequency analysis of TMS-evoked EEG oscillations,"
multiscale Lempel-Ziv complexity in depression," (in eng), Annual (in English), Biomedical Signal Processing and Control, vol. 70, p.
International Conference of the IEEE Engineering in Medicine and 102975, Sep 2021.
Biology Society, pp. 4158-4161, 2015. [43] F. Tadel, S. Baillet, J. C. Mosher, D. Pantazis, and R. M. Leahy,
[23] J. I. Aunon, C. D. Mcgillem, and D. G. Childers, "Signal-Processing in "Brainstorm: A User-Friendly Application for MEG/EEG Analysis," (in
Evoked-Potential Research - Averaging and Modeling," (in English), Crc English), Computational Intelligence and Neuroscience, vol. 2011, p.
Critical Reviews in Bioengineering, vol. 5, no. 4, pp. 323-367, 1981. 879716, 2011.
[24] P. Da Pelo, M. De Tommaso, A. Monaco, S. Stramaglia, R. Bellotti, and [44] A. Gramfort, T. Papadopoulo, E. Olivi, and M. Clerc, "OpenMEEG:
S. Tangaro, "Trial latencies estimation of event-related potentials in EEG opensource software for quasistatic bioelectromagnetics," (in English),
by means of genetic algorithms," Journal of Neural Engineering, vol. 15, Biomedical Engineering Online, vol. 9, no. 1, p. 45, Sep 6 2010.
no. 2, p. 026016, Apr 2018. [45] M. I. Dutt and W. Saadeh, "Monitoring Level of Hypnosis Using
[25] A. Arieli, A. Sterkin, A. Grinvald, and A. Aertsen, "Dynamics of ongoing Stationary Wavelet Transform and Singular Value Decomposition
activity: Explanation of the large variability in evoked cortical responses," Entropy With Feedforward Neural Network," IEEE Transactions on
(in English), Science, vol. 273, no. 5283, pp. 1868-1871, Sep 27 1996. Neural Systems and Rehabilitation Engineering, vol. 31, pp. 1963-1973,
[26] M. N. Shadlen and W. T. Newsome, "The variable discharge of cortical 2023.
neurons: Implications for connectivity, computation, and information [46] E. Maris and R. Oostenveld, "Nonparametric statistical testing of EEG-
coding," (in English), Journal of Neuroscience, vol. 18, no. 10, pp. 3870- and MEG-data," Journal of Neuroscience Methods, vol. 164, no. 1, pp.
3896, May 15 1998. 177-190, 2007/08/15/ 2007.
[27] A. Gramfort, R. Keriven, and M. Clerc, "Graph-Based Variability [47] P. Lioumis, D. Kicic, P. Savolainen, J. P. Mäkekä, and S. Kähköen,
Estimation in Single-Trial Event-Related Neural Responses," (in English), "Reproducibility of TMS - Evoked EEG Responses," (in English), Human
IEEE Transactions on Biomedical Engineering, vol. 57, no. 5, pp. 1051- Brain Mapping, vol. 30, no. 4, pp. 1387-1396, Apr 2009.
1061, May 2010. [48] I. Premoli et al., "TMS- EEG Signatures of GABAergic
[28] S. Musall, M. T. Kaufman, A. L. Juavinett, S. Gluf, and A. K. Churchland, Neurotransmission in the Human Cortex," (in English), Journal of
"Single-trial neural dynamics are dominated by richly varied movements," Neuroscience, vol. 34, no. 16, pp. 5603-5612, Apr 16 2014.
(in English), Nature Neuroscience, vol. 22, no. 10, pp. 1677-+, Oct 2019. [49] Y. Noda et al., "Resting-state EEG gamma power and theta- gamma
[29] A. Schurger, I. Sarigiannidis, L. Naccache, J. D. Sitt, and S. Dehaene, coupling enhancement following high- frequency left dorsolateral
"Cortical activity is more stable when sensory stimuli are consciously prefrontal rTMS in patients with depression," (in English), Clinical
perceived," (in eng), no. 1091-6490 (Electronic). Neurophysiology, vol. 128, no. 3, pp. 424-432, Mar 2017.
[30] J. F. Mitchell, K. A. Sundberg, and J. H. Reynolds, "Differential Attention- [50] P. Dhami et al., "Prefrontal Cortical Reactivity and Connectivity Markers
Dependent Response Modulation across Cell Classes in Macaque Visual Distinguish Youth Depression from Healthy Youth," (in English),
Area V4," Neuron, vol. 55, no. 1, pp. 131-141, 2007. Cerebral Cortex, vol. 30, no. 7, pp. 3884-3894, Jul 2020.
[31] S. Liu et al., "Neurophysiological markers of depression detection and [51] P. Dhami et al., "Neurophysiological markers of response to theta burst
severity prediction in first-episode major depressive disorder," (in English), stimulation in youth depression," Depression and Anxiety, vol. 38, no. 2,
Journal of Affective Disorders, vol. 331, pp. 8-16, Jun 15 2023. pp. 172-184, 2021/02/01 2021.
[32] A. H. Kemp, M. A. Gray, R. B. Silberstein, S. M. Armstrong, and P. J. [52] N. Eshel et al., "Global connectivity and local excitability changes
Nathan, "Augmentation of serotonin enhances pleasant and suppresses underlie antidepressant effects of repetitive transcranial magnetic
unpleasant cortical electrophysiological responses to visual emotional stimulation," Neuropsychopharmacology, vol. 45, no. 6, pp. 1018-1025,
stimuli in humans," NeuroImage, vol. 22, no. 3, pp. 1084-1096, 2020/05/01 2020.
2004/07/01/ 2004. [53] M. S. Hoseini, N. C. Wright, J. Xia, W. Clawson, W. Shew, and R. Wessel,
[33] S. Tremblay et al., "Clinical utility and prospective of TMS–EEG," "Dynamics and sources of response variability and its coordination in
Clinical Neurophysiology, vol. 130, no. 5, pp. 802-844, 2019/05/01/ 2019. visual cortex," (in English), Visual Neuroscience, vol. 36, p. E012, Dec 16
[34] K. X. Cao et al., "TMS-EEG: An emerging tool to study the 2019.
neurophysiologic biomarkers of psychiatric disorders," (in English), [54] S. Isomura et al., "Differentiation between major depressive disorder and
Neuropharmacology, vol. 197, p. 108574, Oct 1 2021. bipolar disorder by auditory steady-state responses," J Affect Disord, vol.
[35] E. Kallioniemi, J. Saari, F. Ferreri, and S. Määttä, "TMS-EEG responses 190, pp. 800-806, Jan 15 2016.
across the lifespan: Measurement, methods for characterisation and [55] G. Pfurtscheller and F. H. Lopes da Silva, "Event-related EEG/MEG
identified responses," (in English), Journal of Neuroscience Methods, vol. synchronization and desynchronization: basic principles," Clin
366, p. 109430, Jan 15 2022. Neurophysiol, vol. 110, no. 11, pp. 1842-57, Nov 1999.
[36] R. Sparing, D. Buelte, I. G. Meister, T. Pauš, and G. R. Fink, "Transcranial [56] G. Thut and A. Pascual-Leone, "A Review of Combined TMS-EEG
magnetic stimulation and the challenge of coil placement: A comparison Studies to Characterize Lasting Effects of Repetitive TMS and Assess
of conventional and stereotaxic neuronavigational strategies," Human Their Usefulness in Cognitive and Clinical Neuroscience," (in English),
Brain Mapping, vol. 29, no. 1, pp. 82-96, 2008/01/01 2008. Brain Topography, vol. 22, no. 4, pp. 219-232, Jan 2010.
[37] E. M. Wassermann, "Variation in the response to transcranial magnetic [57] P. J. Fitzgerald and B. O. Watson, "Gamma oscillations as a biomarker for
brain stimulation in the general population," (in English), Clinical major depression: an emerging topic," (in English), Translational
Neurophysiology, vol. 113, no. 7, pp. 1165-1171, Jul 2002. Psychiatry, vol. 8, no. 1, p. 177, Sep 4 2018.
[38] A. Mir-Moghtadaei et al., "Concordance Between BeamF3 and MRI- [58] S. J. Haghighi, M. Komeili, D. Hatzinakos, and H. El Beheiry, "40-Hz
neuronavigated Target Sites for Repetitive Transcranial Magnetic ASSR for Measuring Depth of Anaesthesia During Induction Phase," (in
Stimulation of the Left Dorsolateral Prefrontal Cortex," (in English), Brain English), IEEE Journal of Biomedical and Health Informatics, vol. 22, no.
Stimulation, vol. 8, no. 5, pp. 965-973, Sep-Oct 2015. 6, pp. 1871-1882, Nov 2018.
[39] L. M. Koponen, S. M. Goetz, and A. V. Peterchev, "Double-Containment [59] C. Spironelli et al., "Evidence of language-related left hypofrontality in
Coil With Enhanced Winding Mounting for Transcranial Magnetic Major Depression: An EEG Beta band study," (in English), Scientific
Stimulation With Reduced Acoustic Noise," (in English), IEEE Reports, vol. 10, no. 1, p. 8166, May 18 2020.
Transactions on Biomedical Engineering, vol. 68, no. 7, pp. 2233-2240, [60] Y. He, W. Guo, Z. Ren, J. Liang, S. Liu, and D. Ming, "Gamma auditory
Jul 2021. steady-state response as a promising electrophysiological biomarker for
[40] N. C. Rogasch et al., "Analysing concurrent transcranial magnetic depression: an in vivo study with chronic unpredictable mild stress (CUS)-
stimulation and electroencephalographic data: A review and introduction induced rats," Cerebral Cortex, vol. 33, no. 12, pp. 7741-7753, 2023.
to the open-source TESA software," (in English), Neuroimage, vol. 147, [61] D. A. Pizzagalli, L. A. Peccoralo, R. J. Davidson, and J. D. Cohen,
no. 15, pp. 934-951, Feb 15 2017. "Resting anterior cingulate activity and abnormal responses to errors in
[41] Y. Wang et al., "Application of Fast Perturbational Complexity Index to subjects with elevated depressive symptoms: a 128-channel EEG study,"
the Diagnosis and Prognosis for Disorders of Consciousness," (in English), Hum Brain Mapp, vol. 27, no. 3, pp. 185-201, Mar 2006.
IEEE Transactions on Neural Systems and Rehabilitation Engineering, [62] M. C. Pellicciari, V. Ponzo, C. Caltagirone, and G. Koch, "Restored
vol. 30, pp. 509-518, 2022. Asymmetry of Prefrontal Cortical Oscillatory Activity after Bilateral
Theta Burst Stimulation Treatment in a Patient with Major Depressive

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
 This article has been accepted for publication in IEEE Transactions on Neural Systems and Rehabilitation Engineering. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2024.3486759

10

Disorder: A TMS-EEG Study," Brain Stimulation, vol. 10, no. 1, pp. 147-
149, 2017/01/01/ 2017.
[63] G. Sanacora et al., "Subtype-specific alterations of γ-aminobutyric acid
and glutamate in patients with major depression," (in English), Archives of
General Psychiatry, vol. 61, no. 7, pp. 705-713, Jul 2004.
[64] A. K. Engel and P. Fries, "Beta-band oscillations — signalling the status
quo?," Current Opinion in Neurobiology, vol. 20, no. 2, pp. 156-165, Apr
2010.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/