Content uploaded by Wanze Xie
Author content
All content in this area was uploaded by Wanze Xie on Dec 07, 2021
Content may be subject to copyright.
CHAPTER
A state-of-the-art methodological
review of pediatric EEG 18
Wanze Xie
a,b
and Charles A. Nelson
c,d,e
School of Psychological and Cognitive Sciences, Peking University, China
a
PKU-IDG/McGovern Institute for Brain
Research, Peking University, China
b
Boston Children’s Hospital, Boston, MA, United States
c
Harvard Medical School,
Boston, MA, United States
d
Harvard Graduate School of Education, Cambridge, MA, United States
e
1. Introduction to pediatric EEG
Electroencephalography (EEG) offers an easy-to-use tool to measure brain function in pediatric popu-
lations. It is often the method of choice when measuring brain activity in awake infants and children due
to its superb temporal resolution, low-cost of recordings, and relatively higher tolerance to children’s
movements compared to other brain imaging methods, e.g., functional magnetic resonance imaging
(fMRI). Using EEG allows researchers to monitor brain activity in young children during resting-state
and cognitive tasks to understand the neurodevelopmental origin of human perception, cognition, and
emotion.
Two widely adopted pediatric EEG measures are EEG power and event-related potential (ERP).
The measure of EEG power in different frequency bands provides information regarding oscillatory
activations in the brain at task and rest. ERP reflects time-locked changes in the brain’s electrical ac-
tivity in response to a perceptual or cognitive challenge. Both measures contribute to our understanding
of the neural correlates of cognitive development in various domains, including attention (Richards,
2003a), face and emotion perception (de Haan and Nelson, 1997;Lepp€
anen et al., 2007), action ob-
servation and execution (Marshall and Meltzoff, 2015), language acquisition (Kuhl, 2010), as well
as memory and executive function (Wolfe and Bell, 2004). In addition, they are convenient and valid
measurements to investigate the deviations from normative brain maturation in early experiences, e.g.,
adversities and neurodevelopmental disorders (Gabard-Durnam et al., 2019;Pierce et al., 2019;
Wilkinson et al., 2019;Xie et al., 2019b). Articles and book series that offer comprehensive practical
guides for using these two measures with pediatric populations are also available (Brooker et al., 2019;
de Haan, 2007).
Although EEG power and ERP have contributed significantly to many aspects of our knowledge of
early brain development, both tools provide limited information on the underlying neural generators
(sources) nor the synchrony and connectivity among brain regions. To fill these gaps, recent advances
have been made in adopting state-of-the-art techniques in pediatric EEG research, such as MRI-
compatible EEG source localization and functional connectivity analysis. This chapter elaborates
on these new techniques and their recent applications in the study of brain and cognitive development.
Handbook of Pediatric Brain Imaging. https://doi.org/10.1016/B978-0-12-816633-8.00014-4
Copyright #2021 Elsevier Inc. All rights reserved. 373
Section 2 of the chapter gives an overview of MRI-compatible EEG source localization and introduces
how it has been used along with conventional EEG power and ERP measures to inform neural mech-
anisms. Section 3 reviews cutting-edge methods to calculate the EEG functional connectivity and ex-
plains how they can be incorporated with EEG source localization to estimate connections between
brain regions for pediatric participants.
There are two other characteristics of pediatric EEG/ERP: relatively low signal-to-noise ratio
(SNR) and inconsistency in the selection of target ERP components, their time-window(s) and elec-
trode location(s). While most researchers understand these are “naturally embedded features” in pedi-
atric EEG studies, an alternative EEG approach, the steady-state visual evoked potential (ssVEP), has
gained popularity among pediatric EEG researchers. This approach is purported to increase the SNR
and allow researchers to objectively define the neural markers in their study (Rossion, 2014). Section 4
of this chapter will introduce the advantages and limitations of this method, as well as its recent ap-
plications in pediatric EEG studies.
2. MRI-compatible cortical source localization in pediatric EEG research
EEG signals recorded by electrodes placed on the scalp represent postsynaptic potentials generated by
mass synchronized pyramidal neurons perpendicular to the cortical surface. Source localization or
source analysis is the most often used method to identify underlying cortical generators (sources)
of EEG potentials measured on the scalp. The sources of EEG signals can be modeled as electrical
dipoles in the brain with three different features: position, directionality, and magnitude. The identi-
fication of the position of cortical sources is often obscured because the current generated by them
spreads in all directions in the brain and gets smeared by the skull. To alleviate this problem, techniques
to improve the spatial resolution of source localization have been constantly developed and refined.
These efforts have made EEG a comprehensive and powerful brain imaging tool with reasonable spa-
tial and superior temporal resolution (Michel and Murray, 2012).
Given the aforementioned advantages of using EEG with pediatric populations, MRI-compatible
source localization with high-density EEG recordings has increasingly emerged as one of the top
choices to investigate brain functioning in children. This section gives an overview of EEG source lo-
calization and the importance of using realistic head models in source localization for pediatric popu-
lations, followed by descriptions on recent studies using this technique as a neuroimaging tool to study
brain functions and cognitive development in childhood.
2.1 EEG source localization
There are two major approaches to conduct source localization—equivalent current dipole (ECD) and
distributed source modeling. ECD modeling uses a limited number of electrical dipoles or sources that
are computed with a forward model to explain the distribution of the EEG on the scalp. A forward
model is created with the locations of the electrodes on the scalp and a MRI head model that describes
the tissues in the head and their conductivity values (Hallez et al., 2007). Distributed source modeling
uses an inverse model along with the obtained EEG data to generate a large set of sources distributed
across the brain. The inverse model is the “inverse” of the forward model and used to compute the
position, directionality, and magnitude of the sources given the scalp EEG (He et al., 2018;Michel
et al., 2004).
374 Chapter 18 Cortical source analysis of EEG
ECD source localization. ECD modeling assumes that the electrical potential over the entire scalp
can be explained by a small set of dipoles (He et al., 1987). These hypothetical dipoles can vary in
position, magnitude, and orientation in a 3D space. A forward model needs to be constructed to esti-
mate the electrical activity over the scalp in ECD modeling. The forward model represents the head
geometry and tissue conductivity and delineates how the activation generated by the dipoles propagates
to the scalp; the so called “lead-field” matrix. The estimated electrical activity over the scalp is cal-
culated by applying the forward model to the current dipoles with certain orientations and magnitudes
(i.e., the forward solution). For ECD modeling, the output of this so-called forward solution can be
compared to the actual electrical activity on the scalp, and thus the amount of variance explained
by the selected current dipoles can be calculated (Richards, 2003b;Scherg, 1992). The optimal solution
is gained through the iteration of the forward solution with different parameters (position, orientation,
and magnitude) of the dipoles until the minimal residual variance is obtained (Scherg et al., 1999).
Alternatively, a set of a priori fixed locations can be set based on theoretical specifications of the known
effects, and thus only the parameters of orientation and magnitude are estimated. ECD models are gen-
erally overdetermined because the number of dipole parameters are significantly less than the number
of surface sensors (electrodes, MEG positions) (Michel et al., 2004).
A practical concern associated with ECD modeling is the uncertainty about the number of dipoles
and their locations to be tested by the forward solution. One solution to this problem is to make a priori
assumptions of the number and location of dipoles based on a theoretical rationale from previous re-
search using other neuroimaging tools, e.g., fMRI and positron emission tomography (PET) (Agam
et al., 2011;Foxe et al., 2003;Gao et al., 2019).
Distributed source modeling. In distributed source models, cortical dipoles are distributed over
the entire source space, each with a fixed position. The positions of the dipoles are called the source
space. The source space can be gray matter (GM) voxels derived from a structural MRI, the surface of
the brain, the inner compartment in boundary element methods, or the entire brain volume. A forward
model is also needed for distributed source modeling. For distributed source modeling, the forward
model is combined with the source space to estimate an inverse spatial filter (Grech et al., 2008).
The inverse spatial filter, when multiplied by the observed scalp electrical current distribution, recon-
structs the current density across the entire set of potential source positions.
The computation of the inverse spatial filter is problematic because the inverse of the lead-field
matrix is “underdetermined” due to the number of sources being substantially larger than the number
of surface electrodes. Thus, additional constraints must be imposed to obtain unique and well-posed
linear inverse solutions (Grech et al., 2008;He et al., 2018). The solution to the underdetermined con-
struction of the inverse spatial filter is to constrain the solution by some set of mathematical or quan-
titative procedures (Grech et al., 2008;Michel and He, 2012). There are a few widely adopted solutions
in distributed source modeling, such as minimum norm estimation (MNE), standardized low-resolution
electromagnetic tomography (sLORETA), exact low-resolution electromagnetic tomography (eLOR-
ETA), and beamforming techniques. Detailed description of these methods can be found elsewhere
(Grech et al., 2008;Green and McDonald, 2009;Hallez et al., 2007;He et al., 2018).
2.2 Realistic head models for source localization in pediatric populations
The accuracy of source localization increases as the anatomical features of issues inside a head are
more realistically represented and discriminated in the head model (Michel et al., 2004;Reynolds
and Richards, 2009). Thus, using age-specific realistic head models for source localization is
3752 MRI-compatible cortical source localization in pediatric EEG research
particularly important for pediatric populations, as there are substantial neuroanatomical changes of the
tissues inside the head over childhood (e.g., synaptic pruning, myelination of axons and neurogenesis),
and the structure of a child brain differs greatly from an adult brain (Phan et al., 2018;Reynolds and
Richards, 2009;Richards and Xie, 2015). For example, the skull conductivity value and thickness are
age dependent (Wendel et al., 2010), such that the skull conductivity value is much higher for infants
than adults (Odabaee et al., 2014). Using adult skull conductivity values for infant EEG data may give
rise to sources that are shallower in the cortex than where they should be.
A significant advance in cortical source analysis with pediatric participants is to use realistic head
models created with individual MRIs or an age-appropriate MRI template (Guy et al., 2016;
H€
am€
al€
ainen et al., 2011;Ortiz-Mantilla et al., 2012;Xie et al., 2017). Although systematic estimation
of skull conductivity for children at different ages has not been conducted, studies have tried to use
higher skull conductivity values for source localization with pediatric EEG data (H€
am€
al€
ainen et al.,
2011;Ortiz-Mantilla et al., 2012;Xie et al., 2018a). There are now age-specific MRI templates avail-
able to the public for research purposes, which can be used to create realistic head models for children
(e.g., the Neurodevelopment MRI Database) (Richards et al., 2016). Since the fontanels and unknitted
skull sutures of an infant head may allow current flow to the scalp unimpeded by the skull, a future
direction is to take these features into account when creating the realistic head models for infants.
2.3 Application of source localization in pediatric EEG research
Source localization of pediatric ERPs. Using cortical source localization techniques, researchers are
able to identify the potential neural generators of ERP components in children. For example, the N290
ERP component is regarded as the precursor to the adult face-sensitive component, the N170 compo-
nent (de Haan and Nelson, 1999;Halit et al., 2004). However, the cortical sources of this “infant face-
sensitive component” remained unclear until two recent studies conducted source localization of the
infant N290 component using the eLORETA method with realistic infant MRI models (Guy et al.,
2016;Xie et al., 2019c). Both studies localized the infant N290 component to the fusiform face area
(FFA) including the fusiform gyrus and inferior occipital gyrus, which is consistent with the cortical
sources found for the adult N170 component (Gao et al., 2019). These studies also examined the cor-
tical sources of two other frequently studied infant visual ERP components, the P400 and Nc compo-
nents. These two components were predominantly localized to the posterior cingulate cortex (PCC) and
precuneus regions, which highlights the important role that the P400 and Nc play in infant arousal and
attention systems (Richards, 2003a;Xie and Richards, 2016). Source localization with age-appropriate
head models has also been used to study the neural generators of infant auditory ERP components, such
as ERPs in response to syllables and pitch changes (H€
am€
al€
ainen et al., 2011;Ortiz-Mantilla et al.,
2012), as well as well-studied visual evoked potentials (VEPs), e.g., the P1 and N1 ( Jensen et al.,
2019;Xie and Richards, 2017)(Fig. 1).
For older children and adolescents, the multimodal neuroimaging analysis could be conducted by a
combination of MRI-constrained EEG source localization and fMRI techniques. This may provide con-
vergent evidence on the underlying brain mechanisms of cognitive processes (Buzzell et al., 2017). A
recent study examined the development of face-elicited brain activation in 9–12-year-old children
using ERP source localization and fMRI measures (Liu et al., 2019). The authors found that the tomog-
raphy of the ERP sources broadly corresponded with the fMRI activation evoked by the same facial
376 Chapter 18 Cortical source analysis of EEG
stimuli, such that a core face-processing system including the inferior occipito-temporal and parietal
regions was identified by both modalities.
Source localization of pediatric EEG rhythmic oscillations. EEG oscillations in different fre-
quency bands have functional significances for cognitive development in various domains (Bell and
Cuevas, 2012). Cortical source analysis of EEG rhythmic activation helps us to understand the neuro-
developmental origin of different cognitive processes. For instance, the mu rhythm (central alpha) is
proposed to be associated with the human mirror neuron system (Marshall and Meltzoff, 2011). EEG
source localization has been utilized to study the neurodevelopmental origin of the mirror neuron sys-
tem in early childhood. A study by Thorpe and colleagues investigated the development of the mu ac-
tivity during motor execution from childhood to adulthood (Thorpe et al., 2016). Distributed source
analysis localized the mu activation to frontal and parietal regions, including the pre- and postcentral
gyri, as well as the precuneus and inferior parietal lobule. The source locations of the mu rhythm were
found to be consistent across 1- and 4-year-olds and adults. The finding of the posterior parietal regions
being part of the sources generating the mu rhythm is reminiscent of two previous studies examining
the infant mu rhythm using ECD models (Nystrom, 2008;Nystrom et al., 2011).
Prior studies have suggested that EEG rhythmic activity can be manipulated by infant visual atten-
tion. However, the cortical generators of brain rhythmic activity associated with infant attention remain
FIG. 1
Cortical source localization results of the infant N290, P400 and Nc components. Distributed source
modeling was used with eLORETA method. The hot colors (red and yellow) in this figure represent brain
regions showing greater activation in the time windows of the ERP components, which means they are more
likely to be the cortical sources of the ERPs.
This figure is adopted from Xie, W., McCormick, S. A., Westerlund, A., Bowman, L. C., & Nelson, C. A., 2019c. Neural correlates of
facial emotion processing in infancy. Dev. Sci. e12758. https://doi.org/10.1111/desc.12758.
3772 MRI-compatible cortical source localization in pediatric EEG research
unclear. To fill this knowledge gap, a recent study explored the complex patterns of EEG oscillations
and their cortical sources observed during infant sustained attention (Xie et al., 2018b). Cortical source
reconstruction of EEG power in different frequency bands was conducted with the eLORETA method
and realistic head models created for each participant using individual MRIs. Infant sustained attention
was found to be accompanied by increased theta power in the orbitofrontal and ventral temporal areas
and decreased alpha power in brain regions within the default mode network (DMN). The relations
between infant sustained attention and EEG power were not shown in infants at 6 and 8 months but
emerged at 10 months and became well established by 12 months.
3. EEG functional connectivity and brain network measures for children
Progress made in developing novel neuroimaging tools in the past few decades makes it possible to
investigate the dynamic interregional communications inside the brain and their development over
childhood. Although fMRI is a frequently used neuroimaging to study the development of functional
brain networks (Goldenberg and Galvan, 2015), the measure of EEG offers a comparatively inexpen-
sive and easy-to-use alternative to achieve this research goal, particularly when dealing with pediatric
populations (Boersma et al., 2013, 2011;Miskovic et al., 2015). The advantages stemming from the
nature of EEG recording make it a more practical tool to estimate functional connectivity while young
children are performing cognitive tasks or in rest-state while they are awake.
Construction of functional networks requires recording physiological or electrical signals from
multiple spatial locations that can either be brain regions of interest (ROIs) or channels of EEG. Func-
tional connectivity inside a network is typically estimated by analyzing the correlation or coherence
between dynamic signals recorded at multiple locations. There are two widely used methods to define
EEG functional networks—component analyses and seed-based connectivity maps. Component anal-
ysis, such as the independent and principle component analyses (ICA and PCA), highlights the brain
networks (i.e., components) that share variance in EEG time series (Liu et al., 2017). Seed-based con-
nectivity maps are comprised of predefined spatial locations (e.g., ROIs and electrodes) where EEG
signals are synchronized with those in other seeds (Xie et al., 2018a, b). A broader view of EEG func-
tional brain networks can be further pursued using graph theory measures to model the topology of
networks and characterize their overall architecture (Rubinov and Sporns, 2010;Vertes and
Bullmore, 2015).
The following paragraphs introduce methods that have been used to estimate the functional con-
nectivity with EEG recordings. This is followed by a review on recent studies on EEG functional con-
nectivity during task performance and the development of brain functional networks and using resting-
state EEG recordings and graph theory measures. It should be noted that the combination of EEG
source localization and functional connectivity analysis allows researchers to measure functional brain
networks in the source-space (cortical level). Thus, this section also describes a few studies that have
applied this technique to study source-space EEG functional connectivity in children.
3.1 Measures of EEG functional connectivity
A few methods have been developed to calculate functional connectivity among EEG signals. Fre-
quently used methods in the field include coherence, phase-lock value, the imaginary part of the co-
herency (IC), phase lag index (PLI), weighted phase lag index (wPLI), and correlation between power
378 Chapter 18 Cortical source analysis of EEG
envelope. The advantages and disadvantages of these methods have been discussed elsewhere (Bastos
and Schoffelen, 2016). Here, we only briefly describe three of these methods, coherence, IC, and wPLI,
and explain the volume conduction problem in EEG functional connectivity analysis.
Cxy fðÞ¼ |Gxy fðÞ|
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Gxx fðÞ
∗Gxx fðÞ:
p
Coherence (C
xy
) between two signals x(t) and y(t) is defined by the equation above. “f” refers to the
Fourier transformation of the signal in a certain time window. “Gxy” stands for the cross-spectral den-
sity (CSD) of the two signals “x” and “y” (e.g., two channels or brain regions) obtained from the Fourier
frequency analysis. “Gxx” and “Gyy” represent the power-spectral density (PSD) for signals “x” and
“y”. It can be seen from the equation that coherence is a measure of the linear relationship between two
signals and is an absolute value ranging between 0 and 1.
Coherence used to be a widely accepted method to estimate EEG functional connectivity and has
been adopted in studies on EEG functional connectivity in pediatric populations (Cuevas et al., 2012;
Thatcher et al., 1987). However, coherence between EEG electrodes can be easily contaminated by
spurious or false positive connections due to the volume conduction or field spread problem (Nolte
et al., 2004;Nunez et al., 1997). The volume conduction problem refers to the fact that the distance
between source generators and EEG electrodes and the tissues that the currents flow through would
lead to a mix of the currents generated by multiple sources. Additionally, there is spatial blurring effect
of the skull on the distribution of EEG signals. Thus, rhythmic oscillations generated by a neuronal
source may be picked up by multiple EEG sensors, particularly when they are close to each other,
which in turn gives rise to unrealistically high coherence among these electrodes.
New methods have been developed to reduce the effects of volume conduction and the field spread
on the estimation of functional connectivity between EEG sensors. The IC is one method that estimates
the functional connectivity between two signals using the imaginary part of coherency (Nolte et al.,
2004). Fourier transform of time series results in complex quantities that include a real part and an
imaginary part. The imaginary part of coherency is only sensitive to the synchronization between
two signals that are time-lagged to each other, i.e., it discards the contributions of 0
°
phase difference
between the signals to their connectivity (Nolte et al., 2004). This is because there should be 0
°
phase
difference (i.e., no time lag) between the spurious connections generated by the same sources, as elec-
trical transmission in the brain is instantaneous. Thus, IC outperforms coherence in measuring the real
interaction (connectivity) between two signals (Nolte et al., 2004).
wPLI is another recently developed method that is more resistant to the volume conduction problem
than the measurement of coherence. wPLI is an extension of the PLI method. PLI estimates to what
extent the phase leads or lags between two signals based on the imaginary part of the CSD of the two
signals (Stam et al., 2007). A problem associated with PLI is that its estimation of the phase leads and
lags can be impacted by noise perturbations in the signals that could possibly have near zero phase
difference (Vinck et al., 2011). wPLI was designed to solve this issue by weighting the phase differ-
ences according to the magnitude of the leads and lags so that phase differences around zero would only
have a marginal contribution to the wPLI value (Vinck et al., 2011).
Graph theory. Functional networks can be described as graphs that are composed of nodes and
edges. The structure of a graph is typically described as a list of nodes and edges. This structure
can be conveniently organized as a matrix termed as an adjacency matrix, which is generated with
the outputs from EEG functional connectivity methods described mentioned above. An adjacency ma-
trix is an N$Nmatrix representing the overall functional network with N$(N%1)/2 unique connec-
tions. Nodes are the components in the graph or matrix that represent the EEG electrodes or ROIs, while
3793 EEG functional connectivity and brain network measures for children
edges are the pairwise correlations/connectivity between the nodes in a functional network (i.e., graph;
see (Bullmore and Sporns, 2009)).
The architecture (topology) of a functional brain network represented by the adjacency matrix can
be qualified and quantified by graph theory measures. Widely used graph theory measures include but
are not limited to path length, clustering coefficient, degree centrality, betweenness centrality, network
hubs, and small-worldness. The definition of these graph theory measures can be found in Chapter 14 of
this book and other review articles (Bullmore and Sporns, 2009;Chu-Shore et al., 2011;Power et al.,
2010;Rubinov and Sporns, 2010).
3.2 EEG functional connectivity analysis in pediatric populations
Development of functional brain networks.There is growing interest in using EEG to examine the
development of brain networks in pediatric populations because of the easy application of EEG and
its tolerance to movement compared to fMRI (Boersma et al., 2011;Smit et al., 2011). The results from
these EEG studies have shown comparable findings on the development of brain functional networks
with those from fMRI studies. Overall, there were changes in both integration and segregation of in-
formation processing in children’s resting-state functional networks measured with EEG recordings.
The examination of connectivity between EEG oscillations using graph theory provides insights
into the changes in the electrophysiological dynamics within functional networks. A longitudinal study
conducted by Boersma et al. (2011) recorded resting-state eyes-closed EEG oscillations from children
at 5 and 7 years of age. Synchronization likelihood (SL) represents the co-oscillation between EEG
signals. Boersma et al. (2011) calculated the SL in three frequency bands (theta: 4–6 Hz, alpha: 6–
11 Hz, and beta: 11–25 Hz) between each pair of electrodes to obtain SL-weighted graphs. The mean
SL over all pairs of electrodes was found to decrease from 5 to 7 years of age. Boersma et al. (2011)
interpreted this finding as reflecting the pruning of unused synapses and the preservation of strong con-
nections, which in turn might result in more cost-effective networks. To test this hypothesis, the authors
calculated the mean normalized clustering coefficient and path length to characterize a network orga-
nization. They found that the average clustering coefficient increased from 5 to 7 years of age in the
alpha rhythm and the path length increased during this age in all three frequency bands. These findings
were interpreted as indicating a shift from random to more organized functional networks during the
development of the brain. Evidence on brain network development also originates from other studies
that have investigated the development of EEG functional connectivity and network topology through-
out the lifespan (Miskovic et al., 2015;Smit et al., 2012, 2011).
Although the findings of functional connectivity between signals in EEG electrodes have shed light
on the development of brain functional networks, limited information about the underlying connections
between brain regions could be inferred from scalp recorded EEG data. The effects of volume conduc-
tion on spurious connections are reduced when analyzing the functional connectivity between recon-
structed cortical source activities (Schoffelen and Gross, 2009). Functional connectivity between brain
ROIs can be estimated after the source localization of EEG time-series in electrodes (Bathelt et al.,
2013;Hillebrand et al., 2012;Xie et al., 2019a). The combination of EEG connectivity and source lo-
calization techniques provides a practical way to examine the development of cortical networks in pe-
diatric populations.
Bathelt et al. (2013) conducted a cortical source analysis of EEG recordings with head models cre-
ated from age-specific MRI templates, and then examined functional connectivity between localized
380 Chapter 18 Cortical source analysis of EEG
activation in cortical regions. The authors found an increase of the node degree, clustering coefficient,
and betweenness centrality of functional brain networks with age (Bathelt et al., 2013), which was in
alignment with previous fMRI research (Power et al., 2010). Bathelt and colleagues also applied ei-
genvalue decomposition to obtain functional brain modules that are separate networks comprising
nodes (i.e., ROIs) that are richly connected within than across the module. The connections within brain
modules remain unchanged but the inter-hemispheric connections between modules increased between
2 and 6 years of age.
EEG functional connectivity during task performance in children.There is accumulating evi-
dence supporting the idea that specific cognitive functions are likely to be carried out by multiple inter-
acting brain areas, which highlights the need for understanding brain-behavior relations from a network
perspective. EEG functional connectivity has been leveraged to investigate neural mechanisms under-
lying child behaviors in various cognitive tasks on attention (Xie et al., 2019a, b, c), error monitoring
(Buzzell et al., 2019), action execution and observation (Debnath et al., 2019), etc. For example, a re-
cent study examined EEG connectivity in the Mu rhythm during action execution and observation in 9-
month-old infants’ (Debnath et al., 2019). EEG functional connectivity in the mu (alpha) frequency
band was estimated with wPLI and found to be elevated between the central and occipital electrodes
during infants’ observation of hand movements, which supports the hypothesis that a distinct functional
connection between the mirror neuron and attention systems might emerge during infants’ observation
of human action.
Studying dynamic interregional communications during different attentional states is critical to un-
derstand infants’ improved performance and elevated ERP responses in cognitive tasks observed dur-
ing sustained attention (Xie and Richards, 2016, 2017). To investigate functional connectivity in brain
networks of awake infants, a group of researchers has recently developed a pipeline to estimate func-
tional connectivity in the cortical source space (Fig. 2;Xie et al., 2019a, b, c). This pipeline includes
cortical source reconstruction of EEG recordings with age-appropriate MRIs, estimating brain func-
tional connectivity between brain regions, and applying graph theory measures to examine the overall
architecture of brain networks. Using this pipeline, the authors were able to study how functional con-
nectivity in brain networks, such as the dorsal and ventral attention, default mode (DMN), and somato-
sensory networks, changes across different attentional states. Their results revealed that infant heart-
rated defined sustained attention is associated with attenuated connectivity in the alpha band within the
dorsal attention network and the DMN, as well as distinct network organization and efficiency indi-
cated by graph theory measures (Xie et al., 2019a, b, c). These findings provide direct evidence for
the important role that the dorsal attention network and the DMN have in infant visual attention.
EEG network analysis in children at-risk for atypical development.Understanding the patterns of
functional connectivity in children at-risk for neurodevelopmental disorders carries important clinical
potential and may inform how brain networks are disrupted among these children. For example, there is
support for the proposal that there are disruptions in EEG functional connectivity in the alpha band in
infancy that are associated with later diagnosis of autism spectrum disorder (ASD) (Haartsen et al.,
2019;Orekhova et al., 2014). Both studies have shown that higher functional connectivity in the alpha
band at 14 months is associated with greater severity of restricted and repetitive behaviors at 36 months
in children who met criteria for ASD. EEG functional connectivity has also been regarded as a fruitful
source of potential biomarkers for attention deficit hyperactivity disorder (ADHD). A recent study has
found an altered thalamo-cortical EEG connectivity profile in children with ADHD compared to
healthy controls in alpha, beta, and gamma frequency bands, and using these EEG connectivity features
3813 EEG functional connectivity and brain network measures for children
and machine learning techniques, the two groups can be classified/separated with high accuracy
(Muthuraman et al., 2019). In another study, graph theory was applied to EEG connectivity matrices
to determine whether network topology is different between ADHD children and typically developing
controls (Ahmadlou et al., 2012). Decreased path length and increased clustering coefficient were
found in children with ADHD compared to the control group. This atypical network topology might
reflect more randomly organized brain networks with higher cost for functional connections in ADHD
children.
The easy-to-use characteristic of EEG connectivity measures can also be harnessed to study how
deviations in early experience, e.g., being exposed to early adversities, could shape brain networks, and
affect neurocognitive outcomes in low-resource settings. To this end, a group of researchers have con-
ducted a study examining the mechanistic pathways by which growth faltering in early childhood im-
pacts future cognitive outcomes (Xie et al., 2019a). These investigators explored brain functional
connectivity as a mediator of the effects of growth faltering on cognitive outcomes in a sample of impo-
verished, urban-dwelling, Bangladeshi children. This longitudinal study shows that whole-brain func-
tional connectivity in the source space is associated with future cognitive function and, perhaps more
importantly, that brain functional connectivity in theta and beta bands at 36 months mediates the re-
lation between growth faltering and children’s IQ 1 year later. These findings provide the first evidence
of network connectivity as a neural pathway by which exposure to early adversity derails cognitive
development in children living in low-resource settings. Another study using similar EEG connectivity
and source localization methods examined the effect of prematurity (preterm birth) on early cortical
network connectivity in newborns (Tokariev et al., 2019). The most prominent markers of prematurity,
compared to healthy control infants, were found in functional connections involving the frontal re-
gions, and these connections were correlated with newborn neurological performance.
4. ssVEP in pediatric EEG research
The steady-state visual evoked potential (ssVEP) is a type of evoked potential elicited by presenting a
sequence of stimuli at a fixed rate, which can be obtained from scalp-recorded EEG signals. This ap-
proach has recently been adopted in studies on the neural bases of child visual perception. Compared to
the conventional EEG and ERP measures, the major advantages of this approach include high SNR and
objectively defined neural markers (frequencies) of interest (Rossion, 2014). These strengths have led
to growing interest in using ssVEP to inform the neurodevelopmental origin of a few cognitive func-
tions, such as visual attention (Christodoulou et al., 2018;Robertson et al., 2012) and face perception
and recognition (de Heering and Rossion, 2015;Farzin et al., 2012). The current section gives a brief
overview of the ssVEP approach, up-to-date examples on its application in children, as well as the lim-
itations of this method in pediatric EEG research.
4.1 Introduction to ssVEP and its advantages
The ssVEPs are cortical responses generated at frequencies that are exact integer multiples of the stim-
ulus presentation rate (Regan, 1989). The ssVEP method was developed by Regan (1966) in his study
on human brain responses to low-level visual stimuli. In that study, Regan examined phase-locked neu-
ral responses to modulated light (luminance flicker) at different target frequency bins with respect to
3834 ssVEP in pediatric EEG research
adjacencies frequencies (noise level). These steady-state or periodic brain responses to visual stimuli
were identified predominately in central occipital electrodes and were found to be independent of the
alpha activity that is a neural index of attention (Regan, 1966).
The frequency of stimulus presentation in an ssVEP paradigm is typically decided depending on the
cognitive process of interest. There is a hypothesis that the stimulation frequency that generates the
highest amplitude of the response is inversely related to the time needed for sufficiently processing
the stimulus (Norcia et al., 2015). For instance, Regan (1966) found that the maximal brain response
to the luminance flicker that evoked the process of low-level visual cues was at 10 Hz. However, a
lower frequency (i.e., slower speed of presentation) may be needed for high-level cognitive processes
(e.g., language processing and face perception). Indeed, a frequency rate of &6 Hz for stimulus pre-
sentation has been used in studies on adult face perception (Rossion and Boremanse, 2011;Rossion
et al., 2012), and presentation rate might need to be even slower for children (Farzin et al., 2012).
The ssVEP approach has a number of advantages compared to traditional ERP designs. One advan-
tage of the ssVEP method is that this kind of frequency-tagging paradigm allows experimenters to ob-
jectively target the periodic visual response(s) of interest at a predefined frequency. For example, a
6 Hz stimulation sequence will provide a robust basic response at 6 Hz, and a target stimulus presented
at the every 5th stimulation will elicit a distinct activation at 1.2 Hz (5/6 ¼1.2). As a result, the basic
and target responses are both confined to predefined frequencies of interest. A second advantage is that
the fast-periodic stimulation sequence used in ssVEP experiments provides data with high SNR be-
cause a large number of stimuli can be presented in a relatively short period of time, e.g., 20–30 s.
Moreover, noise in the data is distributed over the entire spectrum, and thus only a tiny fraction of
the noise will be mixed with the target frequency bin of interest (Regan, 1989). These advantages make
it possible to obtain a reliable periodic response at the individual level through only a few minutes of
stimulus presentation. Therefore, the ssVEP approach has opened a new avenue for studying the neural
correlates of cognitive processes in pediatric populations who have limited attention span.
4.2 Recent application of ssVEP
The ssVEP approach has been leveraged to study the neural correlates of face perception and catego-
rization in childhood. Farzin et al. (2012) examined whether there was differential cortical activity un-
derlying the structural encoding of human faces compared to objects in 4- and 6-month-old infants. The
visual stimuli were presented at a frequency of 6 Hz (i.e., six images/s) and consisted of alternating
scrambled and intact images (faces or objects) at a frequency of 3 Hz (i.e., three images/s for each cat-
egory). This paradigm isolated brain responses to the structural differences between intact and scram-
bled images at the first harmonics (i.e., 3 Hz) from responses to the common local features in intact and
scrambled images at the image update rate (i.e., 6 Hz). The authors found that infants showed signif-
icant ssVEP responses at 3 Hz for both faces and objects, meaning that they were able to detect the
changes in the global structural information in these stimuli. Perhaps more importantly, the amplitude
of the ssVEP response at 3 Hz was greater for faces than objects and the scalp distribution of this re-
sponse was different between the two types of stimuli. These findings suggest that infants aged between
4 and 6 months already show more distinct neural responses to faces than objects, which is consistent
with the existing infant ERP literature (de Haan and Nelson, 1999;Peykarjou and Hoehl, 2013;Xie and
Richards, 2016).
384 Chapter 18 Cortical source analysis of EEG
de Heering and Rossion (2015) further investigated the rapid categorization of natural human faces
from non-face objects in infants at this age (i.e., between 4 and 6 months) (de Heering and Rossion,
2015). During a stimulation train of 20 s, infants were presented with different images of objects at a
fixed rate of 6 Hz, with face images varying in their size, viewpoint, luminance, gender, expression,
and age, embedded at every fifth image (i.e., 1.2 Hz). The idea behind this paradigm is that if the in-
fant’s visual system discriminates between the two categories, their ssVEP response should be distinct
at 1.2 Hz compared to the noise level. The response at 6 Hz and its harmonics was found to be located in
the middle posterior regions, reflecting a general visual response to all stimuli (including faces),
whereas the 1.2 Hz response was located in the occipito-temporal regions (de Heering and Rossion,
2015). This kind of ssVEP design was recently adopted by a study testing whether infants’ ability
to categorize faces from nonface objects could be boosted by their mother’s odor (Leleu et al., 2019).
The strengths of the ssVEP approach also make it a useful tool for studying the development of
attentional dynamics of visual search over childhood. Robertson et al. (2012) tested whether 12-
week-old infants’ ssVEP could be modulated by infant overt attention. Twelve-week-old infants were
presented with a toy duck at the center of the screen in the first experiment. There were LED lights
installed in the duck flicking at 8 Hz. The duck rotated back and forth for 2 s after infants fixated
at it in the experimental condition, while it did not rotate in the control condition. The assumption
was that rotating (moving) objects would increase infants’ attention. The authors analyzed brain re-
sponses to the flicking duck at 8 Hz before and after the onset of its rotation and found that the ssVEP
amplitude at 8 Hz increased significantly after the offset of the rotation in the experimental condition
only, which indicates the effect of overt attention on infant ssVEPs. The effect of covert orienting on
infant brain activation has also been studied using the ssVEP approach. In a recent experiment con-
ducted by Christodoulou et al. (2018),6$6 and 4 $4 checkerboards were presented respectively
in infants’ peripheral fields (left and right) and an attractor was presented in the center. They were
flicking at a frequency of 6 or 12 Hz. The 6 $6 checkerboard is supposed to elicit more visual attention
from young infants than the 4 $4 checkerboard. Christodoulou et al. (2018) found that the amplitude
of the ssVEP to the 6 $6 checkerboard was greater than that to the 4 $4 checkerboard regardless of
the flicking frequency. This finding provides convergent evidence for the effect of covert orienting on
infants’ visual processing of a peripheral stimulus.
4.3 Limitations
Although the ssVEP approach has gradually become an alternative tool to investigate the neurodeve-
lopmental origins of cognitive functions, a couple of limitations should be kept in mind. First, the fre-
quency of stimulus presentation, i.e., how fast the stimulation sequence is presented, should be
determined based on the age of participants and the cognitive functions examined. For example, a fre-
quency rate of &6 Hz for stimulus presentation has been adopted from studies on adult face perception
in recent child ssVEP research; however, whether this rate is adequate for young children to process and
discriminate different types of faces remains unclear given their low visual acuity and processing ef-
ficiency. Future research may consider testing a sweep ssVEP response to justify the presentation rate/
speed for various kinds of visual stimuli and different age groups, as what has been done for adults
(Alonso-Prieto et al., 2013).
The ssVEP approach is limited in offering precise temporal and spatial information of brain acti-
vation due to the periodic nature of steady-state stimuli, especially when they were presented at high
3854 ssVEP in pediatric EEG research
rates (Norcia et al., 2015). The Fourier transform of EEG time series eliminates the temporal informa-
tion of the cognitive processes triggered by the stimulation sequence. Although researchers have
attempted to analyze the ERP components evoked by the fast and periodic stimulus presentation
(Yan et al., 2019), directly relating results of ssVEP responses to the temporal evolution of ERPs is
still challenging. The spatial resolution of ssVEP is also restricted due to the volume conduction issue.
However, source localization of child ssVEP responses with age-appropriate MRIs should be advo-
cated in future research, as the validity of source localization relies on the SNR in the EEG data.
The high SNR in ssVEP offers an advantage of incorporating source localization into ssVEP analysis.
5. Conclusion
In this chapter, we reviewed recent progress made in pediatric EEG research and state-of-the-art EEG
methods that have rendered EEG a top brain imaging tool to study the child brain. The straightforward
compatibility of EEG with other brain imaging techniques, especially structural MRI, leads to the
growing interest in using EEG source localization with age-appropriate MRI models to investigate
the neural mechanisms underlying cognitive functions in children. Cutting-edge methods in EEG func-
tional connectivity analysis now allow researchers to study the development of brain networks in in-
fancy and how the trajectory could be derailed by neurodevelopmental disorders or exposure to early
adversities. Moreover, alternative paradigms like ssVEP have recently been adopted to increase the
SNR and objectively define neural markers of interests in pediatric EEG research. In spite of the many
issues still to overcome in this filed, all the efforts and advances reviewed in this chapter have dem-
onstrated how quickly the field is growing and the promising future of pediatric EEG research.
References
Agam, Y., Hamalainen, M.S., Lee, A.K., Dyckman, K.A., Friedman, J.S., Isom, M., Manoach, D.S., 2011. Mul-
timodal neuroimaging dissociates hemodynamic and electrophysiological correlates of error processing. Proc.
Natl. Acad. Sci. U. S. A. 108 (42), 17556–17561. https://doi.org/10.1073/pnas.1103475108.
Ahmadlou, M., Adeli, H., Adeli, A., 2012. Graph theoretical analysis of organization of functional brain networks
in ADHD. Clin. EEG Neurosci. 43 (1), 5–13. https://doi.org/10.1177/1550059411428555.
Alonso-Prieto, E., Belle, G.V., Liu-Shuang, J., Norcia, A.M., Rossion, B., 2013. The 6 Hz fundamental stimulation
frequency rate for individual face discrimination in the right occipito-temporal cortex. Neuropsychologia 51
(13), 2863–2875. https://doi.org/10.1016/j.neuropsychologia.2013.08.018.
Bastos, A.M., Schoffelen, J.M., 2016. A tutorial review of functional connectivity analysis methods and their in-
terpretational pitfalls. Front. Syst. Neurosci. 9. https://doi.org/10.3389/Fnsys.2015.00175, 175.
Bathelt, J., O’Reilly, H., Clayden, J.D., Cross, J.H., de Haan, M., 2013. Functional brain network organisation of
children between 2 and 5 years derived from reconstructed activity of cortical sources of high-density EEG
recordings. NeuroImage 82, 595–604. https://doi.org/10.1016/j.neuroimage.2013.06.003.
Bell, M.A., Cuevas, K., 2012. Using EEG to study cognitive development: issues and practices. J. Cogn. Dev. 13
(3), 281–294. https://doi.org/10.1080/15248372.2012.691143.
Boersma, M., Smit, D.J., de Bie, H.M., Van Baal, G.C., Boomsma, D.I., de Geus, E.J., Stam, C.J., 2011. Network
analysis of resting state EEG in the developing young brain: structure comes with maturation. Hum. Brain
Mapp. 32 (3), 413–425. https://doi.org/10.1002/hbm.21030.
386 Chapter 18 Cortical source analysis of EEG
Boersma, M., Smit, D.J., Boomsma, D.I., De Geus, E.J., Delemarre-van de Waal, H.A., Stam, C.J., 2013. Growing
trees in child brains: graph theoretical analysis of electroencephalography-derived minimum spanning tree in
5- and 7-year-old children reflects brain maturation. Brain Connect. 3 (1), 50–60. https://doi.org/10.1089/
brain.2012.0106.
Brooker, R.J., Bates, J.E., Buss, K.A., Canen, M.J., Dennis-Tiwary, T.A., Gatzke-Kopp, L.M., Schmidt, L.A.,
2019. Conducting event-related potential (ERP) research with young children. J. Psychophysiol., 1–
22. https://doi.org/10.1027/0269-8803/a000243.
Bullmore, E., Sporns, O., 2009. Complex brain networks: graph theoretical analysis of structural and functional
systems. Nat. Rev. Neurosci. 10 (3), 186–198. https://doi.org/10.1038/nrn2575.
Buzzell, G.A., Richards, J.E., White, L.K., Barker, T.V., Pine, D.S., Fox, N.A., 2017. Development of the error-
monitoring system from ages 9-35: unique insight provided by MRI-constrained source localization of EEG.
NeuroImage 157, 13–26. https://doi.org/10.1016/j.neuroimage.2017.05.045.
Buzzell, G.A., Barker, T.V., Troller-Renfree, S.V., Bernat, E.M., Bowers, M.E., Morales, S., Fox, N.A., 2019.
Adolescent cognitive control, theta oscillations, and social observation. NeuroImage 198, 13–30. https://
doi.org/10.1016/j.neuroimage.2019.04.077.
Christodoulou, J., Leland, D.S., Moore, D.S., 2018. Overt and covert attention in infants revealed using steady-
state visually evoked potentials. Dev. Psychol. 54 (5), 803–815. https://doi.org/10.1037/dev0000486.
Chu-Shore, C.J., Kramer, M.A., Bianchi, M.T., Caviness, V.S., Cash, S.S., 2011. Network analysis: applications
for the developing brain. J. Child Neurol. 26 (4), 488–500. https://doi.org/10.1177/0883073810385345.
Cuevas, K., Swingler, M.M., Bell, M.A., Marcovitch, S., Calkins, S.D., 2012. Measures of frontal functioning and
the emergence of inhibitory control processes at 10 months of age. Dev. Cogn. Neurosci. 2 (2), 235–
243. https://doi.org/10.1016/j.dcn.2012.01.002.
de Haan, M. (Ed.), 2007. Infant EEG and Event-Related Potentials. Psychological Press, New York.
de Haan, M., Nelson, C.A., 1997. Recognition of the mother’s face by six-month-old infants: a neurobehavioral
study. Child Dev. 68 (2), 187–210. https://doi.org/10.2307/1131845.
de Haan, M., Nelson, C.A., 1999. Brain activity differentiates face and object processing in 6-month-old infants.
Dev. Psychol. 35 (4), 1113–1121.
de Heering, A., Rossion, B., 2015. Rapid categorization of natural face images in the infant right hemisphere. elife
4. https://doi.org/10.7554/eLife.06564, e06564.
Debnath, R., Salo, V.C., Buzzell, G.A., Yoo, K.H., Fox, N.A., 2019. Mu rhythm desynchronization is specific to
action execution and observation: evidence from time-frequency and connectivity analysis. NeuroImage 184,
496–507. https://doi.org/10.1016/j.neuroimage.2018.09.053.
Farzin, F., Hou, C., Norcia, A.M., 2012. Piecing it together: infants’ neural responses to face and object structure.
J. Vis. 12 (13), 6. https://doi.org/10.1167/12.13.6.
Foxe, J.J., McCourt, M.E., Javitt, D.C., 2003. Right hemisphere control of visuospatial attention: line-bisection
judgments evaluated with high-density electrical mapping and source analysis. NeuroImage 19 (3), 710–726.
Gabard-Durnam, L.J., Wilkinson, C., Kapur, K., Tager-Flusberg, H., Levin, A.R., Nelson, C.A., 2019. Longitu-
dinal EEG power in the first postnatal year differentiates autism outcomes. Nat. Commun. 10 (1), 4188. https://
doi.org/10.1038/s41467-019-12202-9.
Gao, C., Conte, S., Richards, J.E., Xie, W., Hanayik, T., 2019. The neural sources of N170: understanding timing of
activation in face-selective areas. Psychophysiology. https://doi.org/10.1111/psyp.13336, e13336.
Goldenberg, D., Galvan, A., 2015. The use of functional and effective connectivity techniques to understand the
developing brain. Dev. Cogn. Neurosci. 12, 155–164. https://doi.org/10.1016/j.dcn.2015.01.011.
Grech, R., Cassar, T., Muscat, J., Camilleri, K.P., Fabri, S.G., Zervakis, M., Vanrumste, B., 2008. Review on solv-
ing the inverse problem in EEG source analysis. J. Neuroeng. Rehabil. 5, 25. https://doi.org/10.1186/1743-
0003-5-25.
Green, J.J., McDonald, J.J., 2009. A practical guide to beamformer source reconstruction for EEG. In: Handy, T.C.
(Ed.), Brain Signal Analysis: Advances in Neuroelectric and Neuromagnetic Methods. The MIT Press, Cam-
bridge, MA, pp. 76–98.
387References
Guy, M.W., Zieber, N., Richards, J.E., 2016. The cortical development of specialized face processing in infancy.
Child Dev. https://doi.org/10.1111/cdev.12543.
Haartsen, R., Jones, E.J.H., Orekhova, E.V., Charman, T., Johnson, M.H., Team, B., 2019. Functional EEG con-
nectivity in infants associates with later restricted and repetitive behaviours in autism; a replication study.
Transl. Psychiatry 9 (1), 66. https://doi.org/10.1038/s41398-019-0380-2.
Halit, H., Csibra, G., Volein, A., Johnson, M.H., 2004. Face-sensitive cortical processing in early infancy. J. Child
Psychol. Psychiatry 45 (7), 1228–1234. https://doi.org/10.1111/j.1469-7610.2004.00321.x.
Hallez, H., Vanrumste, B., Grech, R., Muscat, J., De Clercq, W., Vergult, A., Lemahieu, I., 2007. Review on solv-
ing the forward problem in EEG source analysis. J. Neuroeng. Rehabil. 4, 46. https://doi.org/10.1186/1743-
0003-4-46.
H€
am€
al€
ainen, J.A., Ortiz-Mantilla, S., Benasich, A.A., 2011. Source localization of event-related potentials to pitch
change mapped onto age-appropriate MRIs at 6 months of age. NeuroImage 54 (3), 1910–1918. https://doi.org/
10.1016/j.neuroimage.2010.10.016.
He, B., Musha, T., Okamoto, Y., Homma, S., Nakajima, Y., Sato, T., 1987. Electric dipole tracing in the brain by
means of the boundary element method and its accuracy. IEEE Trans. Biomed. Eng. 34 (6), 406–414.
He, B., Sohrabpour, A., Brown, E., Liu, Z., 2018. Electrophysiological source imaging: a noninvasive window to
brain dynamics. Annu. Rev. Biomed. Eng. 20, 171–196. https://doi.org/10.1146/annurev-bioeng-062117-
120853.
Hillebrand, A., Barnes, G.R., Bosboom, J.L., Berendse, H.W., Stam, C.J., 2012. Frequency-dependent functional
connectivity within resting-state networks: an atlas-based MEG beamformer solution. NeuroImage 59 (4),
3909–3921. https://doi.org/10.1016/j.neuroimage.2011.11.005.
Jensen, S.K.G., Kumar, S., Xie, W., Tofail, F., Haque, R., Petri, W.A., Nelson, C.A., 2019. Neural correlates of
early adversity among Bangladeshi infants. Sci. Rep. 9 (1), 3507. https://doi.org/10.1038/s41598-019-39242-
x.
Kuhl, P.K., 2010. Brain mechanisms in early language acquisition. Neuron 67 (5), 713–727. https://doi.org/
10.1016/j.neuron.2010.08.038.
Leleu, A., Rekow, D., Poncet, F., Schaal, B., Durand, K., Rossion, B., Baudouin, J.Y., 2019. Maternal odor shapes
rapid face categorization in the infant brain. Dev. Sci. https://doi.org/10.1111/desc.12877, e12877.
Lepp€
anen, J.M., Moulson, M.C., Vogel-Farley, V.K., Nelson, C.A., 2007. An ERP study of emotional face pro-
cessing in the adult and infant brain. Child Dev. 78 (1), 232–245. https://doi.org/10.1111/j.1467-
8624.2007.00994.x.
Liu, Q., Farahibozorg, S., Porcaro, C., Wenderoth, N., Mantini, D., 2017. Detecting large-scale networks in the
human brain using high-density electroencephalography. Hum. Brain Mapp. https://doi.org/10.1002/
hbm.23688.
Liu, P., Bai, X., Perez-Edgar, K.E., 2019. Integrating high-density ERP and fMRI measures of face-elicited brain
activity in 9-12-year-old children: an ERP source localization study. NeuroImage 184, 599–608. https://doi.
org/10.1016/j.neuroimage.2018.09.070.
Marshall, P.J., Meltzoff, A.N., 2011. Neural mirroring systems: exploring the EEG mu rhythm in human infancy.
Dev. Cogn. Neurosci. 1 (2), 110–123. https://doi.org/10.1016/j.dcn.2010.09.001.
Marshall, P.J., Meltzoff, A.N., 2015. Body maps in the infant brain. Trends Cogn. Sci. 19 (9), 499–505. https://doi.
org/10.1016/j.tics.2015.06.012.
Michel, C., He, B., 2012. EEG mapping and source imaging. In: Schomer, D., Lopes da Silva, F.H. (Eds.), Nie-
dermeyer’s Electroencephalography. Lippincott Williams & Wilkins, Philadelphia, PA, pp. 1179–1202.
Michel, C.M., Murray, M.M., 2012. Towards the utilization of EEG as a brain imaging tool. NeuroImage 61 (2),
371–385. https://doi.org/10.1016/j.neuroimage.2011.12.039.
Michel, C.M., Murray, M.M., Lantz, G., Gonzalez, S., Spinelli, L., Grave de Peralta, R., 2004. EEG source im-
aging. Clin. Neurophysiol. 115 (10), 2195–2222. https://doi.org/10.1016/j.clinph.2004.06.001.
388 Chapter 18 Cortical source analysis of EEG
Miskovic, V., Ma, X., Chou, C.A., Fan, M., Owens, M., Sayama, H., Gibb, B.E., 2015. Developmental changes in
spontaneous electrocortical activity and network organization from early to late childhood. NeuroImage 118,
237–247. https://doi.org/10.1016/j.neuroimage.2015.06.013.
Muthuraman, M., Moliadze, V., Boecher, L., Siemann, J., Freitag, C.M., Groppa, S., Siniatchkin, M., 2019. Mul-
timodal alterations of directed connectivity profiles in patients with attention-deficit/hyperactivity disorders.
Sci. Rep. 9 (1), 20028. https://doi.org/10.1038/s41598-019-56398-8.
Nolte, G., Bai, O., Wheaton, L., Mari, Z., Vorbach, S., Hallett, M., 2004. Identifying true brain interaction from
EEG data using the imaginary part of coherency. Clin. Neurophysiol. 115 (10), 2292–2307. https://doi.org/
10.1016/j.clinph.2004.04.029.
Norcia, A.M., Appelbaum, L.G., Ales, J.M., Cottereau, B.R., Rossion, B., 2015. The steady-state visual evoked
potential in vision research: a review. J. Vis. 15 (6), 4. https://doi.org/10.1167/15.6.4.
Nunez, P.L., Srinivasan, R., Westdorp, A.F., Wijesinghe, R.S., Tucker, D.M., Silberstein, R.B., Cadusch, P.J.,
1997. EEG coherency. 1. Statistics, reference electrode, volume conduction, Laplacians, cortical imaging,
and interpretation at multiple scales. Electroencephalogr. Clin. Neurophysiol. 103 (5), 499–515. https://doi.
org/10.1016/S0013-4694(97)00066-7.
Nystrom, P., 2008. The infant mirror neuron system studied with high density EEG. Soc. Neurosci. 3 (3–4), 334–
347. https://doi.org/10.1080/17470910701563665.
Nystrom, P., Ljunghammar, T., Rosander, K., von Hofsten, C., 2011. Using mu rhythm desynchronization to mea-
sure mirror neuron activity in infants. Dev. Sci. 14 (2), 327–335.
Odabaee, M., Tokariev, A., Layeghy, S., Mesbah, M., Colditz, P.B., Ramon, C., Vanhatalo, S., 2014. Neonatal
EEG at scalp is focal and implies high skull conductivity in realistic neonatal head models. NeuroImage
96, 73–80. https://doi.org/10.1016/j.neuroimage.2014.04.007.
Orekhova, E.V., Elsabbagh, M., Jones, E.J., Dawson, G., Charman, T., Johnson, M.H., Team, B., 2014. EEG
hyper-connectivity in high-risk infants is associated with later autism. J. Neurodev. Disord. 6 (1),
40. https://doi.org/10.1186/1866-1955-6-40.
Ortiz-Mantilla, S., Hamalainen, J.A., Benasich, A.A., 2012. Time course of ERP generators to syllables in infants:
a source localization study using age-appropriate brain templates. NeuroImage 59 (4), 3275–3287. https://doi.
org/10.1016/j.neuroimage.2011.11.048.
Peykarjou, S., Hoehl, S., 2013. Three-month-olds’ brain responses to upright and inverted faces and cars. Dev.
Neuropsychol. 38 (4), 272–280. https://doi.org/10.1080/87565641.2013.786719.
Phan, T.V., Smeets, D., Talcott, J.B., Vandermosten, M., 2018. Processing of structural neuroimaging data in
young children: bridging the gap between current practice and state-of-the-art methods. Dev. Cogn. Neurosci.
33, 206–223. https://doi.org/10.1016/j.dcn.2017.08.009.
Pierce, L.J., Thompson, B.L., Gharib, A., Schlueter, L., Reilly, E., Valdes, V., Nelson, C.A., 2019. Association of
perceived maternal stress during the perinatal period with electroencephalography patterns in 2-month-old in-
fants. JAMA Pediatr. https://doi.org/10.1001/jamapediatrics.2019.0492.
Power, J.D., Fair, D.A., Schlaggar, B.L., Petersen, S.E., 2010. The development of human functional brain net-
works. Neuron 67 (5), 735–748. https://doi.org/10.1016/j.neuron.2010.08.017.
Regan, D., 1966. Some characteristics of average steady-state and transient responses evoked by modulated light.
Electroencephalogr. Clin. Neurophysiol. 20, 238–248.
Regan, D., 1989. Human Brain Electrophysiology: Evoked Potentials and Evoked Magnetic Fields in Science and
Medicine. Elsevier, Amsterdam.
Reynolds, G.D., Richards, J.E., 2009. Cortical source localization of infant cognition. Dev. Neuropsychol. 34 (3),
312–329. https://doi.org/10.1080/87565640902801890. PII 910996962.
Richards, J.E., 2003a. Attention affects the recognition of briefly presented visual stimuli in infants: an ERP study.
Dev. Sci. 6 (3), 312–328. https://doi.org/10.1111/1467-7687.00287.
389References
Richards, J.E., 2003b. Cortical sources of event-related potentials in the prosaccade and antisaccade task. Psycho-
physiology 40 (6), 878–894.
Richards, J.E., Xie, W., 2015. Brains for all the ages: structural neurodevelopment in infants and children from a
life-span perspective. In: Benson, J. (Ed.), Advances in Child Development and Behavior. vol. 48. Elsevier,
Philadephia, PA, pp. 1–52.
Richards, J.E., Sanchez, C., Phillips-Meek, M., Xie, W., 2016. A database of age-appropriate average MRI tem-
plates. NeuroImage 124, 1254–1259. https://doi.org/10.1016/j.neuroimage.2015.04.055.
Robertson, S.S., Watamura, S.E., Wilbourn, M.P., 2012. Attentional dynamics of infant visual foraging. Proc. Natl.
Acad. Sci. U. S. A. 109 (28), 11460–11464. https://doi.org/10.1073/pnas.1203482109.
Rossion, B., 2014. Understanding face perception by means of human electrophysiology. Trends Cogn. Sci. 18 (6),
310–318. https://doi.org/10.1016/j.tics.2014.02.013.
Rossion, B., Boremanse, A., 2011. Robust sensitivity to facial identity in the right human occipito-temporal cortex
as revealed by steady-state visual-evoked potentials. J. Vis. 11 (2). https://doi.org/10.1167/11.2.16.
Rossion, B., Prieto, E.A., Boremanse, A., Kuefner, D., Van Belle, G., 2012. A steady-state visual evoked potential
approach to individual face perception: effect of inversion, contrast-reversal and temporal dynamics. Neuro-
Image 63 (3), 1585–1600. https://doi.org/10.1016/j.neuroimage.2012.08.033.
Rubinov, M., Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. Neuro-
Image 52 (3), 1059–1069. https://doi.org/10.1016/j.neuroimage.2009.10.003.
Scherg, M., 1992. Functional imaging and localization of electromagnetic brain activity. Brain Topogr. 5 (2),
103–111.
Scherg, M., Bast, T., Berg, P., 1999. Multiple source analysis of interictal spikes: goals, requirements, and clinical
value. J. Clin. Neurophysiol. 16 (3), 214–224.
Schoffelen, J.M., Gross, J., 2009. Source connectivity analysis with MEG and EEG. Hum. Brain Mapp. 30 (6),
1857–1865. https://doi.org/10.1002/hbm.20745.
Smit, D.J., de Geus, E.J., van de Nieuwenhuijzen, M.E., van Beijsterveldt, C.E., van Baal, G.C., Mansvelder, H.D.,
Linkenkaer-Hansen, K., 2011. Scale-free modulation of resting-state neuronal oscillations reflects prolonged
brain maturation in humans. J. Neurosci. 31 (37), 13128–13136. https://doi.org/10.1523/JNEUROSCI.1678-
11.2011.
Smit, D.J., Boersma, M., Schnack, H.G., Micheloyannis, S., Boomsma, D.I., Hulshoff Pol, H.E., de Geus, E.J.,
2012. The brain matures with stronger functional connectivity and decreased randomness of its network. PLoS
One 7 (5). https://doi.org/10.1371/journal.pone.0036896, e36896.
Stam, C.J., Nolte, G., Daffertshofer, A., 2007. Phase lag index: assessment of functional connectivity from multi
channel EEG and MEG with diminished bias from common sources. Hum. Brain Mapp. 28 (11), 1178–
1193. https://doi.org/10.1002/hbm.20346.
Thatcher, R.W., Walker, R.A., Giudice, S., 1987. Human cerebral hemispheres develop at different rates and ages.
Science 236 (4805), 1110–1113.
Thorpe, S.G., Cannon, E.N., Fox, N.A., 2016. Spectral and source structural development of mu and alpha rhythms
from infancy through adulthood. Clin. Neurophysiol. 127 (1), 254–269. https://doi.org/10.1016/j.
clinph.2015.03.004.
Tokariev, A., Roberts, J.A., Zalesky, A., Zhao, X., Vanhatalo, S., Breakspear, M., Cocchi, L., 2019. Large-scale
brain modes reorganize between infant sleep states and carry prognostic information for preterms. Nat. Com-
mun. 10 (1). https://doi.org/10.1038/s41467-019-10467-8.
Vertes, P.E., Bullmore, E.T., 2015. Annual research review: growth connectomics—the organization and reorga-
nization of brain networks during normal and abnormal development. J. Child Psychol. Psychiatry 56 (3), 299–
320. https://doi.org/10.1111/jcpp.12365.
Vinck, M., Oostenveld, R., van Wingerden, M., Battaglia, F., Pennartz, C.M., 2011. An improved index of phase-
synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size
bias. NeuroImage 55 (4), 1548–1565. https://doi.org/10.1016/j.neuroimage.2011.01.055.
390 Chapter 18 Cortical source analysis of EEG
Wendel, K., Vaisanen, J., Seemann, G., Hyttinen, J., Malmivuo, J., 2010. The influence of age and skull conduc-
tivity on surface and subdermal bipolar EEG leads. Comput. Intell. Neurosci. 2010, 397272. https://doi.org/
10.1155/2010/397272.
Wilkinson, C.L., Levin, A.R., Gabard-Durnam, L.J., Tager-Flusberg, H., Nelson, C.A., 2019. Reduced frontal
gamma power at 24 months is associated with better expressive language in toddlers at risk for autism. Autism
Res. 12 (8), 1211–1224. https://doi.org/10.1002/aur.2131.
Wolfe, C.D., Bell, M.A., 2004. Working memory and inhibitory control in early childhood: contributions from
physiology, temperament, and language. Dev. Psychobiol. 44 (1), 68–83. https://doi.org/10.1002/dev.10152.
Xie, W., Richards, J.E., 2016. Effects of interstimulus intervals on behavioral, heart rate, and event-related poten-
tial indices of infant engagement and sustained attention. Psychophysiology 53 (8), 1128–1142. https://doi.org/
10.1111/psyp.12670.
Xie, W., Richards, J.E., 2017. The relation between infant covert orienting, sustained attention and brain activity.
Brain Topogr. 30 (2), 198–219. https://doi.org/10.1007/s10548-016-0505-3.
Xie, W., Mallin, B.M., Richards, J.E., 2017. Development of infant sustained attention and its relation to EEG
oscillations: an EEG and cortical source analysis study. Dev. Sci. https://doi.org/10.1111/desc.12562.
Xie, W., Jensen, K.G., Wade, M., Kumar, S., Westerlund, A., Kakon, S.H., Nelson, C.A., 2018a. Child growth
predicts brain functional connectivity and future cognitive outcomes in urban Bangladeshi children exposed
to early adversities. bioRxiv. https://doi.org/10.1101/447722, 447722.
Xie, W., Mallin, B.M., Richards, J.E., 2018b. Development of brain functional connectivity and its relation to in-
fant sustained attention in the first year of life. Dev. Sci. https://doi.org/10.1111/desc.12703, e12703.
Xie, W., Jensen, S.K.G., Wade, M., Kumar, S., Westerlund, A., Kakon, S.H., Nelson, C.A., 2019a. Growth fal-
tering is associated with altered brain functional connectivity and cognitive outcomes in urban Bangladeshi
children exposed to early adversity. BMC Med. 17 (1), 199. https://doi.org/10.1186/s12916-019-1431-5.
Xie, W., Kumar, S., Kakon, S.H., Haque, R., Petri, W.A., Nelson, C.A., 2019b. Chronic inflammation is associated
with neural responses to faces in Bangladeshi children. NeuroImage 202, 116110. https://doi.org/10.1016/j.
neuroimage.2019.116110.
Xie, W., McCormick, S.A., Westerlund, A., Bowman, L.C., Nelson, C.A., 2019c. Neural correlates of facial emo-
tion processing in infancy. Dev. Sci. https://doi.org/10.1111/desc.12758, e12758.
Yan, X., Liu-Shuang, J., Rossion, B., 2019. Effect of face-related task on rapid individual face discrimination.
Neuropsychologia 129, 236–245. https://doi.org/10.1016/j.neuropsychologia.2019.04.002.
391References
