This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number

OptoNet II: An Advanced MATLAB-based

Toolbox for Functional Cortical Connectivity
Analysis with Surrogate Tests Using fNIRS
Gihyoun Lee1,2, Ji-Su Park3, Jungsoo Lee1, Jinuk Kim2, Young-Jin Jung4,*, and Yun-Hee
Kim1,2,*
1
Department of Physical and Rehabilitation Medicine, Center for Prevention and Rehabilitation, Heart Vascular Stroke Institute, Samsung Medical
Center, Sungkyunkwan University School of Medicine, Seoul, 06351 Korea
2
Department of Health Sciences and Technology, SAIHST, Sungkyunkwan University, Seoul, 06351 Korea
3
Advanced Human Resource Development Project Group for Health Care in Aging Friendly Industry, Dongseo University, Busan, 47011 Korea
4
Department of Radiological Science, Health Sciences Division, DongSeo University, Busan, 47011 Korea

Corresponding author: Yun-Hee Kim (e-mail: [email protected]) and Young-Jin Jung (e-mail: [email protected]).

*The authors contributed equally to this project as co-corresponding authors

ABSTRACT Cortical connectivity analysis is a widely used method for understanding the causes of
neurological disorders and related brain mechanisms. Although there exist numerous activity analysis
toolboxes for functional near-infrared spectroscopy (fNIRS), there are only a few cortical connectivity
analysis toolboxes. In 2019, we released a MATALB toolbox named OptoNet, which has helped
researchers to analyze brain networks using fNIRS. In this study, we developed an advanced MATLAB
toolbox, named OptoNet II, to add new features that overcome the shortcomings of OptoNet. With these
new features, OptoNet II can efficiently analyze cortical connectivity according to brain region using any
fNIRS channel sets and can present the results of two connectivity analyses with auto-thresholding based on
surrogate tests. To evaluate the efficacy of the new functions, the finger-tapping task experiment was
carried out before and after transcranial direct current stimulation (tDCS) in the primary motor area.
OptoNet II can efficiently show the effects of tDCS on functional brain region connectivity, which has been
difficult to confirm by conventional methods. In this paper, we propose the OptoNet II as a useful and
efficient toolbox for researchers who want to perform cortical connectivity analysis using fNIRS.

KEYWORDS Brain network analysis, Brain phase synchronization, Cortical hemodynamic signals,
Cortical connectivity, Functional near-infrared spectroscopy.

I. INTRODUCTION regions, and have also shown agreement with regards to

Functional neural connectivity plays an important role in quantifying the level of synchronization [2].
advanced cognitive processes, and has thus drawn increasing fNIRS is a noninvasive method used to measure
attention from researchers over the past decades [1-6]. hemodynamic brain signals based on the absorption of near-
Various measures, such as phase synchronization index (PSI) infrared light, with wavelengths in the range of 650 nm to
[7], mutual information [8], partial directed coherence [9], 950 nm, transmitted through the intact skull [11]. fNIRS
frequency ratio [10], and mean phase coherence [9], have monitors variations in regional cerebral blood flow and
been applied to quantify the functional connectivity of estimates hemodynamic signals, which are highly correlated
different brain units using multichannel neural signals, to the blood-oxygenation-level-dependent (BOLD) signal
including electroencephalography (EEG), functional outputs in fMRI [12]. The important advantages of fNIRS are
magnetic resonance imaging (fMRI), and functional near- its low cost, portability, and the potential to extend research
infrared spectroscopy (fNIRS) [5]. These functional to a wider range of environments than many other neuro-
connectivity measurements have shown similar global imaging systems, such as fMRI, EEG, and
functional connectivity patterns across studies, with some magnetoencephalography (MEG) [13].
differences between study results for particular cortical General linear modeling (GLM) is one of the most
extensively used models to represent data in a linear

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

combination form, and constitutes a standard method for functional areas of the brain, and therefore analyze cortical
analyzing fMRI data. Indeed, many statistical analysis connectivity according to functional region. Additionally,
toolboxes have been developed for fNIRS based on the GLM. OptoNet II features an auto-threshold function based on
However, GLM-based analysis methods often fail to surrogate tests, which can provide significance tests if the
adequately analyze brain functions because of artifacts in the original signal has its property randomized among the
fNIRS measurements. The artifacts exist for various reasons, surrogate data [10]. Detailed descriptions and examples of
such as subject movements, blood pressure variations, and OptoNet II are provided in the following sections. We tested
instrumental instabilities [14-19]. Recently, various OptoNet II using 64-bit Windows 10 installed on Intel i5 and
connection and causality estimation methods for functional i7 personal computer systems.
brain network analysis have been developed, and have
demonstrated their utility in cognitive neuroscience and II. Methods
neurological clinical studies [20]. Many brain network The procedures of OptoNet II can be roughly divided into the
estimation methods have been applied to numerous following steps: signal processing, brain region model setting,
functional neuroimaging modalities, such as EEG, local field and connectivity analysis. In the fNIRS signal processing
potential, intracranial EEG, MEG, fMRI, and fNIRS. step, loading fNIRS data, selecting the fNIRS epoch, and
The most popular tools currently available for fNIRS setting the analysis duration are performed. The connectivity
analysis are hypergeometric optimization of Motif processing step consists of loading the standard head model,
enrichment (HOMER) from Harvard Medical School [21] setting the brain regions, and executing connectivity analysis.
and near-infrared spectroscopy statistical parametric mapping The GUI of the toolbox is optimized for the latest version of
(NIRS–SPM) from the Korea Advanced Institute of Science the MATLAB (ver. R2020a) (MathWorks, Natick, MA,
and Technology (KAIST) [15]. HOMER (available at USA) and GeForce (Nvidia, Santa Clara, CA, USA) graphics
http://www.nmr.mgh.harvard.edu/PMI/) calculates individual
hemodynamic responses using ordinary least-squares linear
deconvolution, while NIRS–SPM (available at
http://bisp.kaist.ac.kr/NIRS-SPM) applies the SPM method,
which refers to the construction and assessment of spatially
extended statistical processes used to test hypotheses about
functional imaging data [3, 15]. However, these analysis
tools cannot estimate functional brain connectivity, and they
are difficult for new users to learn. Therefore, prior to the
release of OptoNet, there was no software program for the
analysis of functional brain networks using fNIRS available
for free and which could be used by unskilled users [3]. In
2019, our research team released a Windows-based graphical
user interface (GUI) MATLAB toolbox named OptoNet with
the aim of providing unexperienced users with an easy way
to analyze cortical networks based on 3-dimensional (3D)
finite element analysis (FEA) [22, 23]. However, OptoNet
can only analyze cortical networks for each channel of any
given fNIRS system; thus, if the system has many channels,
the results of the cortical network analysis can be too
complex to check certain networks. Furthermore, OptoNet
uses a manual threshold setting, which can only show those
networks over the user-determined threshold value. This
manual threshold setting can cause the objectivity and
reliability of the results to be poor.
In this paper, we introduce an advanced version of
OptoNet, named OptoNet II, which is also a Windows-based
MATLAB toolbox. The primary differences between
OptoNet II and OptoNet are the cortical connectivity analysis
according to the functional brain region, the auto threshold
method that uses a surrogate test, and the representation
function of the connectivity difference.
In contrast to the previous version of OptoNet, users of
FIGURE 1. Sequential OptoNet II
OptoNet II can freely set all fNIRS channels to fit the

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

card series. Figure 1 shows the overall sequential steps and shown in Figure 3 A. Figure 3 B shows the “Region
GUI of OptoNet II. Next, detailed descriptions for each step Analysis” section. If the on/off checkbox is checked, the
will be presented. region analysis mode is activated, while if it remains
unchecked, OptoNet II is essentially in fNIRS channel-based
A. SIGNAL PROCESSING STEP
The signal processing step is performed in the signal
processing panel of OptoNet II. First, the measured fNIRS
data can be entered into the “NIRS Data Load” section
indicated in Figure 2 A. OptoNet II supports the .NIR file
format, which can be easily converted from the .mat file
format of MATLAB. OptoNet II also provides an NIRX
converter, as shown in Figure 2 B, which can easily convert
fNIRS data to .NIR files. The loaded fNIRS signals are
plotted, as shown in Figure 2 C, and the plotting type and
fNIRS epoch used to analyze connectivity can be set in the
“Signal Analysis Section” as shown in Figure 2 D. The
processed fNIRS signals are refreshed in the location shown
in Figure 2 C; users do not need to execute this procedure
more than once unless the fNIRS data or epoch are changed.

B. BRAIN REGION MODEL SETTING STEP

As seen in Figure 3, the connectivity processing step is
performed in the connectivity processing panel. Users can
FIGURE 2. The signal processing panel of OptoNet II. (A) “NIRS
select between the standard and the individual head and Data Load” section, (B) NIRX converter, (C) “NIRS Signal Plot”
cortex models in the “Load Head and Cortex Model” section section, and (D) “NIRS Signal Analysis” section.

FIGURE 3. OptoNet II connectivity processing panel. (A) section for selection of the head and cortex models, (B) brain
region analysis section, (B) section to input number of fNIRS channels used for each functional brain region, (C) head
model showing representation of connectivity mapping, (E) section for threshold selection, connectivity type, and
execution of connectivity analysis, and (F) button for analyzing differences in connectivity between datasets.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

analysis mode, similar to the original OptoNet. Users can set each cortical source to create a surrogate data set. In order to
up brain region modeling according to fNIRS channels by shuffle fNIRS signals, the phase and amplitude were
inputting information from the fNIRS channels into the randomized in the Fourier transformed fNIRS signals to
“OptoNet Region Model Setting” section seen in Figure 3 C. preserve the power spectra feature of measured fNIRS
The standard brain region model is separated by 15 regions signals with less distortion. Network analysis was estimated
according to the functional regions of the brain, including the from the surrogate data set by performing the same
medial pre-frontal (MPF), the left and right frontal (Lt. Fr calculating process for 10,000 or more iterations (10,000
and Rt. Fr), the left and right primary motor cortex (Lt. M1 times is a default value in the software). In this way, the
and Rt. M1), supplementary motor area (SMA), the left and empirical distribution for a given network estimator can be
right pre-motor cortex (Lt. PM and Rt. PM), the left and right created. From the distributions, null hypotheses are provided:
temporal lobe (Lt. Tm and Rt. Tm), the left and right sensory “the independently and randomly generated surrogate data
cortex (Lt. Sn and Rt. Sn), the left and right parietal (Lt. Pr set has no interactions between generated fNIRS signals”.
and Rt. Pr), and the occipital cortex (Occ). From these distributions, significant levels for estimated
Once entered, the number of fNIRS channels for each networks were determined from each empirical distribution
functional region can be saved and used again later. The [2, 27]. The surrogate tests used the auto-threshold method
saved brain region model can be loaded by using the button that was added in this new version of OptoNet II, and which
in Figure 3 B, and then the brain regions that will be can be activated using the actions shown in Figure 3 G. To
analyzed are represented in Figure 3 F. If the user wants to select significant networks, the toolbox follows a rank-order
make a custom brain region model that has a different test [28]. First, a residual probability α of false rejection is
number of functional brain regions from the standard region, selected corresponding to a level of significance ((1 − α) ×
the user decides the positions of the region maker on the head 100%) [25]. Then, for a one-tailed test, which can look for
model using the position setter in Figure 3 D. Then, custom small prediction errors, surrogate sequences (M = K / α − 1)
fNIRS channels are entered as many as the number of the are generated, where K is a positive integer. For a two-tailed
custom regions using the custom region setter that repeats test, which can go both ways for time asymmetry, the
pop-up until pressing the save button in Figure 3 E. The surrogates (M = 2K / α − 1 ) are generated, resulting in a
fNIRS channels that are set to the same functional brain probability α that the data gives either one of the K smallest
region are grouped into that region, and connectivity analysis or largest values [25]. Using more surrogates can increase the
is performed according to the set brain region model. discrimination performance [25], and 10,000 surrogate
samples were used in OptoNet II. If a network has a higher
C. CONNECTIVITY ANALYSIS STEP surrogate than the level of significance, it will be represented
One of the biggest problems that exists when estimating as a significant network.
brain connectivity is determining whether corresponding The method used for the connectivity analysis can be
brain region pairs are significantly connected or not with selected from one of following: correlation [29, 30],
high confidence. The previous version of OptoNet only coherence [31], frequency ratio [32], phase locking value
offered the manual threshold setting, which could only (PLV) [33], and is executed as shown in Figure 3 G. After
represent connections over the manually set threshold; analysis, only significant connectivity is represented in
therefore, it could not determine whether there was Figure 3 F by auto-thresholding based on the surrogate tests.
significant connectivity between regions. In order to The resulting values for all estimated connectivities,
overcome this problem, the surrogate data can provide including insignificant connectivities, are saved in a work
significance tests on whether the original signal has the folder as a .mat file. If the user wishes to compare two
property randomized among the surrogate data [24, 25]. connectivity result files (e.g., rest state and task state), the
Therefore, an artificial surrogate test was adopted in this differences in connectivity can be estimated by using the
version of OptoNet II. button shown in Figure 3 H. The difference in the
Surrogate methods usually produce artificial data by connectivities is calculated through a simple subtraction from
randomizing the property to be tested while mimicking other the two estimated connectivity results, and then the
properties (e.g., the spectra) of the original signal as much as significant connectivities of the difference are selected using
possible; the more properties that are preserved, the stricter the surrogate test again. After that program is executed, the
the generated surrogate data [26]. The surrogate time series two results files are continuously loaded and the differences
has the same rank sequence as the Gaussian time series, but in connectivity between the two files are calculated. Then,
the samples of the surrogate time series all come from the the significant differences in connectivity are graphically
measured fNIRS signal. This means that the surrogate series represented on the head model shown in Figure 3 F.
is a rank-shuffled version of the fNIRS signal, with the
temporal structure destroyed but the distribution, mean, and III. Experimental Results
variance all preserved [26]. In other words, the surrogate test We performed an experiment that highlights the improved
randomly and independently shuffles fNIRS signals from features of OptoNet II over the original version. The

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

experimental task paradigm consisted of the finger tapping obtain a baseline fNIRS signal. The finger-tapping task was
task using the left hand. performed for 20 seconds with sequentially provided visual
In this experiment, finger tapping is performed stimulation that permitted a finger press in the same position
sequentially, and participants must accurately tap their finger as the star mark (★). And then, a subject took a rest for 20
when prompted by a sign displayed on the monitor screen seconds and repeated these blocks five times pre- and post-
shown in the right picture in Figure 4 A. Twenty-three adult tDCS.
volunteers over 19 years old who were right-handed were
recruited. All subjects were healthy without a history of brain B. tDCS SYSTEM
injury, neurological, or psychiatric diseases. Study protocols As can be seen in Figure 4 A, the transcranial direct current
were approved by the Samsung Medical Center (SMC) stimulation (tDCS) electrodes were placed on the scalp, with
Institutional Review Board (IRB), and all subjects gave the anode electrode overlying the right M1 region, which is
written informed consent prior to each experiment. the area related to the finger-tapping task of the left-hand,
that was connected to the cathode electrode overlying the left
A. fNIRS SYSTEM M1 region. Stimulation was applied using a DC-
fNIRS data acquisition was performed with an fNIRS brain STIMULATOR (NeuroConn, neuroCare Group, GmbH,
imaging system (NIRScout 24-24, NIRx Medizintechnik Berlin, Germany). The DC current was initially increased in
GmbH, Berlin, Germany). This instrument has a laser near- a ramp-like fashion over several seconds (~10 s) until
infrared light source with four wavelengths at 685 nm, 780 reaching 1 mA, and stimulation was maintained for a total of
nm, 808 nm, and 830 nm. The arrangements of the optode 30 min. DC currents were turned off slowly over a few
structure and fNIRS channels are shown in Figure 4 A, with seconds, out of the field of view of the patients, a procedure
24 NIR sources, 24 NIR detectors, and 81 fNIRS channels. that does not elicit perceived sensations [34].
As can be seen in Figure 4 A, the fNIRS channels cover the
whole brain cortex in order to analyze cortical connectivity,
and it shows the experimental conditions used during fNIRS. C. FUNCTIONAL CONNECTIVITY ANALYSIS
fNIRS data were obtained during the pre-sequential finger- The sequential finger-tapping task was performed before and
tapping task and post-sequential finger tapping task, which after tDCS as shown in Figure 4 B. The experimental block
have the task block before and after tDCS according to the design for the finger tapping task consists of 20-sec task
experimental design in Figure 4 B. The finger-tapping task blocks alternating with 20-sec rest blocks to prevent finger
paradigm starts with the first one-minute eye closed state to fatigue. Five sets of alternating task and rest blocks were

FIGURE 4. Experimental conditions (A) structure of fNIRS and tDCS systems and representation of a subject wearing the
experimental systems, (B) experimental design and finger tapping task paradigm.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

FIGURE 5. Functional cortical connectivity, (A) the result from a previous version of OptoNet, (B) the result of before
tDCS using OptoNet II, (C) the result of after tDCS, and (D) the result of the connectivity difference between (C) and (B)
using OptoNet II.

performed, and a total of 260 seconds of fNIRS data were normalized according to each task block. The fNIRS
acquired, including 1 minute of baseline fNIRS data for channels for each functional brain region were selected
twenty-three participants and a total of 115 task blocks. through the existing toolbox called fNIRS optodes’ location
Estimation of the oxy-hemoglobin (HbO) signal from fNIRS decider (fOLD), which is a first-order approach to bring the
was performed by using the NIRS-SPM toolbox [15] in advanced parcellation methods and meta-analyses acquired
MATLAB. Band-pass filter and normalization were used as from functional magnetic resonance imaging to more
the preprocessing methods for the raw HbO signals. The precisely guide the selection of optode positions for fNIRS
cutoff frequencies were set between 0.02 ~ 0.1 Hz to remove experiments [37]. The decided fNIRS channels were grouped
long-term baseline drifts, high-frequency noise, and cardiac as follows: MPF (CH 19, 77, 78), Lt. Fr (CH 33, 34, 44), Rt.
pulsations [35, 36]. And then, band-passed fNIRS data were Fr (CH 72, 73, 74), Lt. M1 (CH 6, 8, 9, 13), Rt. M1 (CH 11,

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

24, 26, 28), SMA (CH 10, 15, 17, 21), Lt. PM (CH 14, 36), the finger-tapping task with attention to a visual stimulus.
Rt. PM (CH 23, 67), Lt. Tm (CH 42, 43, 44), Rt. Tm (CH 63, The connections between the M1s of both hemispheres and
64, 65), Lt. Sn (CH 2, 3), Rt. Sn (CH 29, 31), Lt. Pr (CH 45, adjacent cortical regions are considered to be the most
46, 47), Rt. Pr (CH 51, 52, 53), and Occ (CH 51). In order to important brain networks related to the motor task used in
avoid signal distortion from differences in the number of this experiment. By using OptoNet II, these networks were
fNIRS channels in each group, the fNIRS signals were visualized by calculating the differences in connectivity
processed with a narrow band-pass filter and normalization, strength between the pre- and post-tDCS conditions
then they were grouped through ensemble averaging. Therefore, the results of this analysis can support that
Functional connectivity was estimated using the PLV [33], changes in connectivity between functional brain regions
and group analysis for all task blocks of all participants was related to the finger tapping task were enhanced following
performed with the use of OptoNet II based on estimations tDCS.
using all trials and all epochs, as shown in Figure 5. The
functional connectivity results obtained using the original IV. Discussion
version of OptoNet are shown in Figure 5 A, and were In 2019, we released a MATLAB toolbox, named OptoNet,
obtained based on fNIRS channels with manual thresholding which can be used to analyze functional cortical networks for
(PLV > 0.8). As results for connectivity between functional fNIRS. It was easy, and simple to use for plotting fNIRS
brain regions cannot be analyzed for significance using the signals and functional cortical connections according to
original OptoNet due to its fNIRS channel-based analysis and fNIRS channels without any anatomical information.
manual thresholding method, the results in Figure 5 A show However, if there are many fNIRS channels, it might be
all connectivity, even insignificant connectivity. Therefore, it difficult for OptoNet to find any given important connection
is not easy to interpret these results, because there are so due to so many connections.
many connections among the fNIRS channels. In order to overcome these issues, we updated OptoNet
functions. The first updated function is functional brain
D. CONNECTIVITY ANALYSIS WITH SURROGATE TEST region-based analysis, which can easily save and load a brain
Figure 5 B, C, and D show the results of connectivity region model after simply entering the number of fNIRS
analyses using the new OptoNet II. Results concisely channels, and it can allow for analysis of the connectivity
presented connectivity between the functional brain regions between functional brain region from fNIRS data. The
and showed only connections that were significant using second new capacity is analysis of connectivity differences,
automatic thresholding with the random surrogate tests of which is a function of loading two different saved
10,000 iterations. Figure 5 B represents the connectivity connectivity results, then representing the different
analysis performed before tDCS, which demonstrated several connectivity by task or state between them. We confirmed
weak connections spread over wide brain areas between the the clearer effect of tDCS through the connectivity difference
frontal, temporal, M1, sensory, parietal, and occipital cortices. between before and after tDCS stimulation. The last new
In particular, the most strong connection was seen between function in OptoNet II is random surrogate tests-based auto-
the occipital and parietal areas which can be interpreted as thresholding, which allows users to automatically identify
neural activities during the process of performing a motor only significant connections. The automatic threshold
task using visual cues [34]. Figure 5 C demonstrates the method shows objective meaningful connections by
results of post-tDCS connectivity analysis that showed fewer statistical significance through random comparisons of more
significant connections. Instead, significant new connections than 10,000 iterations rather than just a cut off based in
were seen between M1, the sensory cortex, and parietal areas simple numerical comparison. Through these updates to
following tDCS. This is a result of the changes in the regions OptoNet II, we are able to provide much better results
related to the motor task and the other associated regions reliability to OptoNet II users.
because the brain networks around M1 and sensory cortices In this work, we developed a freely downloadable
are increased after stimulation and the other networks not MATLAB toolbox, named OptoNet II
associated with motor control are decreased after stimulation. (https://sites.google.com/site/dsucore/free/optonet) for the
This is in line with previous studies, and it is thought that analysis of functional cortical connectivity. Any constructive
cortical networks begin to work efficiently and form essential comments and questions are always welcomed through our e-
connections during a motor task following tDCS [38-40]. mail.
However, to more clearly identify the difference before and For future studies, we will update the OptoNet II toolbox
after tDCS, the difference in connectivity before and after to have even more user-friendly functions. More methods to
stimulation was estimated by using the new toolbox. Figure 5 evaluate functional connectivity networks to be added
D shows the differences in connectivity between Figure 5 C include directed transfer function (DTF), Granger causality,
and B. As shown here, the significantly stronger connections transfer entropy, partial directed coherence, and others as
following tDCS exist between the M1, temporal, parietal, and programming advances continue. We also plan to enable this
occipital regions. These are important connections related to toolbox not only for functional network analysis but also for

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

brain activity analysis so that it can be used in all fNIRS [9] Q. Zhang, Y. Hu, T. Potter, R. Li, M. Quach, and
research fields. Y. Zhang, "Establishing functional brain networks
using a nonlinear partial directed coherence
ACKNOWLEDGMENT method to predict epileptic seizures," Journal of
This study was supported by the Korea Health Technology Neuroscience Methods, vol. 329, p. 108447, 2020.
R&D Project through the Korea Health Industry [10] X. Cheng, Y. Pan, Y. Hu, and Y. Hu,
Development Institute (KHIDI), funded by the Ministry of "Coordination elicits synchronous brain activity
Health & Welfare, Republic of Korea (HI17C1501) and a between co-actors: Frequency ratio matters,"
National Research Foundation of Korea (NRF) grant funded Frontiers in Neuroscience, vol. 13, 2019.
by the Korean government (NRF-2020R1A2C3010304, [11] A. Villringer and U. Dirnagl, "Coupling of brain
NRF-2019R1I1A1A01060660). activity and cerebral blood flow: basis of
functional neuroimaging," Cerebrovascular and
[1] C. L. Yu, H. C. Chen, Z. Y. Yang, and T. L. Chou, brain metabolism reviews, vol. 7, no. 3, pp. 240-
"Multi-time-point analysis: A time course analysis 276, 1995.
with functional near-infrared spectroscopy," Behav [12] Y. Hoshi, "Functional near‐infrared optical
Res Methods, Feb 5 2020. imaging: Utility and limitations in human brain
[2] G. Lee, J.-S. Park, M. L. B. Ortiz, J.-Y. Hong, S.- mapping," Psychophysiology, vol. 40, no. 4, pp.
H. Paik, S. H. Lee, B. M. Kim, and Y.-J. Jung, 511-520, 2003.
"Hemodynamic Activity and Connectivity of the [13] T. Fekete, D. Rubin, J. M. Carlson, and L. R.
Prefrontal Cortex by Using Functional Near- Mujica-Parodi, "The NIRS Analysis Package:
Infrared Spectroscopy during Color-Word noise reduction and statistical inference," PLoS
Interference Test in Korean and English One, vol. 6, no. 9, p. e24322, 2011.
Language," Brain Sciences, vol. 10, no. 8, p. 484, [14] W. Wei, S. Huang, N. Wang, Z. Jin, J. Zhang, and
2020. W. Chen, "A near-infrared spectrometer based on
[3] G. Lee, J.-S. Park, and Y.-J. Jung, "OptoNet: a novel grating light modulators," Sensors, vol. 9, no.
MATLAB-based toolbox for cortical network 4, pp. 3109-3121, 2009.
analyses using functional near-infrared [15] J. C. Ye, S. Tak, K. E. Jang, J. Jung, and J. Jang,
spectroscopy," Optical Engineering, vol. 59, no. "NIRS-SPM: statistical parametric mapping for
06, 2019. near-infrared spectroscopy," Neuroimage, vol. 44,
[4] P. Pinti, I. Tachtsidis, A. Hamilton, J. Hirsch, C. no. 2, pp. 428-47, Jan 15 2009.
Aichelburg, S. Gilbert, and P. W. Burgess, "The [16] K. E. Jang, S. Tak, J. Jung, J. Jang, Y. Jeong, and J.
present and future use of functional near‐infrared C. Ye, "Wavelet minimum description length
spectroscopy (fNIRS) for cognitive neuroscience," detrending for near-infrared spectroscopy," J
Annals of the New York Academy of Sciences, vol. Biomed Opt, vol. 14, no. 3, p. 034004, May-Jun
1464, no. 1, p. 5, 2020. 2009.
[5] J.-W. Park, G. Lee, B.-M. Kim, Y. Chang, and Y.- [17] X. Cui, S. Bray, and A. L. Reiss, "Functional near
J. Jung, "Analytic Signal-Based Causal Network infrared spectroscopy (NIRS) signal improvement
Estimator for Hemodynamic Signal Analysis in the based on negative correlation between oxygenated
Brain," Journal of the Korean Physical Society, and deoxygenated hemoglobin dynamics,"
journal article vol. 74, no. 9, pp. 847-854, May 01 Neuroimage, vol. 49, no. 4, pp. 3039-46, Feb 15
2019. 2010.
[6] S. Sutoko, H. Sato, A. Maki, M. Kiguchi, Y. [18] G. Lee, S. H. Jin, and J. An, "Motion Artifact
Hirabayashi, H. Atsumori, A. Obata, T. Funane, Correction of Multi-Measured Functional Near-
and T. Katura, "Tutorial on platform for optical Infrared Spectroscopy Signals Based on Signal
topography analysis tools," Neurophotonics, vol. 3, Reconstruction Using an Artificial Neural
no. 1, p. 010801, 2016. Network," Sensors (Basel), vol. 18, no. 9, Sep 5
[7] D. Rosenbaum, E. J. Leehr, A. Kroczek, J. A. 2018.
Rubel, I. Int-Veen, K. Deutsch, M. J. Maier, J. [19] G. Lee, S. H. Lee, S. H. Jin, and J. An, "Robust
Hudak, A. J. Fallgatter, and A.-C. Ehlis, "Neuronal functional near infrared spectroscopy denoising
correlates of spider phobia in a combined fNIRS- using multiple wavelet shrinkage based on a
EEG study," Scientific reports, vol. 10, no. 1, pp. hemodynamic response model," Journal of Near
1-14, 2020. Infrared Spectroscopy, vol. 26, no. 2, pp. 79-86,
[8] K. Takagi, "Principles of Mutual Information 2018.
Maximization and Energy Minimization Affect the [20] Y.-J. Jung, K. H. Kim, and C.-H. Im,
Activation Patterns of Large Scale Networks in the "Mathematical Issues in the Inference of Causal
Brain," Frontiers in Computational Neuroscience, Interactions among Multichannel Neural Signals,"
vol. 13, p. 86, 2020.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042808, IEEE Access

Journal of Applied Mathematics, vol. 2012, pp. 1- from Kat landslides (Tokat—Turkey)," Computers
14, 2012. & Geosciences, vol. 35, no. 6, pp. 1125-1138,
[21] T. J. Huppert, S. G. Diamond, M. A. Franceschini, 2009.
and D. A. Boas, "HomER: a review of time-series [33] J. P. Lachaux, E. Rodriguez, J. Martinerie, and F. J.
analysis methods for near-infrared spectroscopy of Varela, "Measuring phase synchrony in brain
the brain," Applied optics, vol. 48, no. 10, pp. signals," Human brain mapping, vol. 8, no. 4, pp.
D280-D298, 2009. 194-208, 1999.
[22] Y.-J. Jung, J.-H. Kim, and C.-H. Im, "COMETS: [34] P. C. Gandiga, F. C. Hummel, and L. G. Cohen,
A MATLAB toolbox for simulating local electric "Transcranial DC stimulation (tDCS): a tool for
fields generated by transcranial direct current double-blind sham-controlled clinical studies in
stimulation (tDCS)," Biomedical Engineering brain stimulation," Clinical neurophysiology, vol.
Letters, vol. 3, no. 1, pp. 39-46, 2013. 117, no. 4, pp. 845-850, 2006.
[23] C. Lee, Y. J. Jung, S. J. Lee, and C. H. Im, [35] F. S. Racz, P. Mukli, Z. Nagy, and A. Eke,
"COMETS2: An advanced MATLAB toolbox for "Increased prefrontal cortex connectivity during
the numerical analysis of electric fields generated cognitive challenge assessed by fNIRS imaging,"
by transcranial direct current stimulation," J Biomedical optics express, vol. 8, no. 8, pp. 3842-
Neurosci Methods, vol. 277, pp. 56-62, Feb 1 2017. 3855, 2017.
[24] M. Palus, "Detecting phase synchronization in [36] S. Ge, Q. Yang, R. Wang, P. Lin, J. Gao, Y. Leng,
noisy systems," Physics Letters A, vol. 235, no. 4, Y. Yang, and H. Wang, "A brain-computer
pp. 341-351, 1997. interface based on a few-channel EEG-fNIRS
[25] T. Schreiber and A. Schmitz, "Surrogate time bimodal system," IEEE Access, vol. 5, pp. 208-218,
series," Physica D: Nonlinear Phenomena, vol. 2017.
142, no. 3-4, pp. 346-382, 2000. [37] G. A. Z. Morais, J. B. Balardin, and J. R. Sato,
[26] J. Sun, X. Hong, and S. Tong, "Phase "fNIRS optodes’ location decider (fOLD): a
synchronization analysis of EEG signals: an toolbox for probe arrangement guided by brain
evaluation based on surrogate tests," IEEE regions-of-interest," Scientific reports, vol. 8, no. 1,
Transactions on Biomedical Engineering, vol. 59, pp. 1-11, 2018.
no. 8, pp. 2254-2263, 2012. [38] R. Polanía, M. A. Nitsche, and W. Paulus,
[27] J. S. Kim, C. H. Im, Y. J. Jung, E. Y. Kim, S. K. "Modulating functional connectivity patterns and
Lee, and C. K. Chung, "Localization and topological functional organization of the human
propagation analysis of ictal source rhythm by brain with transcranial direct current stimulation,"
electrocorticography," NeuroImage, vol. 52, no. 4, Human brain mapping, vol. 32, no. 8, pp. 1236-
pp. 1279-1288, 2010. 1249, 2011.
[28] J. Theiler, S. Eubank, A. Longtin, B. Galdrikian, [39] A. Antal, R. Polania, C. Schmidt-Samoa, P.
and J. D. Farmer, "Testing for nonlinearity in time Dechent, and W. Paulus, "Transcranial direct
series: the method of surrogate data," Physica D: current stimulation over the primary motor cortex
Nonlinear Phenomena, vol. 58, no. 1-4, pp. 77-94, during fMRI," Neuroimage, vol. 55, no. 2, pp.
1992. 590-596, 2011.
[29] A. Aarabi, F. Wallois, and R. Grebe, "Does [40] J. Baudewig, M. A. Nitsche, W. Paulus, and J.
spatiotemporal synchronization of EEG change Frahm, "Regional modulation of BOLD MRI
prior to absence seizures?," Brain Research, vol. responses to human sensorimotor activation by
1188, pp. 207-221, 2008. transcranial direct current stimulation," Magnetic
[30] G. Thut, D. Veniero, V. Romei, C. Miniussi, P. Resonance in Medicine: An Official Journal of the
Schyns, and J. Gross, "Rhythmic TMS causes International Society for Magnetic Resonance in
local entrainment of natural oscillatory Medicine, vol. 45, no. 2, pp. 196-201, 2001.
signatures," Current biology, vol. 21, no. 14, pp.
1176-1185, 2011.
[31] G. Varotto, E. Visani, S. Franceschetti, G.
Sparacino, and F. Panzica, "Spectral and
coherence analysis of EEG during intermittent
photic stimulation in patients with photosensitive
epilepsy," International Journal of
Bioelectromagnetism, vol. 11, no. 4, pp. 189-193,
2009.
[32] I. Yilmaz, "Landslide susceptibility mapping using
frequency ratio, logistic regression, artificial
neural networks and their comparison: a case study

VOLUME XX, 2017 9

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