discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/259959364
CITATIONS READS
11 88
5 authors, including:
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Christos Frantzidis
letting you access and read them immediately. Retrieved on: 17 June 2016
Cognitive and physical training for the elderly: evaluating outcome efficacy
by means of neurophysiological synchronization
PII: S0167-8760(14)00036-1
DOI: doi: 10.1016/j.ijpsycho.2014.01.007
Reference: INTPSY 10761
Please cite this article as: Frantzidis, Christos A., Ladas, Aristea-Kiriaki I., Vivas, Ana
B., Tsolaki, Magda, Bamidis, Panagiotis D., Cognitive and physical training for the
elderly: evaluating outcome efficacy by means of neurophysiological synchronization,
International Journal of Psychophysiology (2014), doi: 10.1016/j.ijpsycho.2014.01.007
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
ACCEPTED MANUSCRIPT
T
Tsolakic and Panagiotis D. Bamidisa, *
R IP
a
Laboratory of Medical Informatics, Medical School, Aristotle University of
Thessaloniki, Greece (christos.frantzidis@gmail.com, bamidis@med.auth.gr)
SC
b
Psychology Department, The University of Sheffield International
Faculty, City College, Thessaloniki, Greece (arladas@seerc.org,
NU
vivas@city.academic.gr)
c
3rd Department of Neurology, Medical School, Aristotle University of Thessaloniki,
MA
Greece (tsolakim@med.auth.gr)
Abstract
Recent neuroscientific research has demonstrated that both healthy and pathological
T
brain. Both compensatory and neurodegenerative mechanisms contribute to
IP
neurophysiological synchronization patterns, which provide a valuable marker for
R
age-related cognitive decline. In this study, we propose that neuroplasticity-based
SC
training may facilitate coherent interaction of distant brain regions and consequently
enhance cognitive performance in elderly people. If this true, this would make
NU
neurophysiological synchronization a valid outcome measure to assess the efficacy of
MA
non-pharmacological interventions to prevent or delay age-related cognitive decline.
computed for all participants so as to quantify the proximity of their individual profile
CE
intervention group as compared to the active control group. It is concluded that the
valid and reliable outcome measure; while the distance-based analysis could provide a
1. Introduction
ACCEPTED MANUSCRIPT
functional changes in the brain (Ewers et al., 2011; Zhang et al., 2012). For instance,
studies involving healthy elderly participants have reported reduced activity in neural
T
networks that are active at rest (resting-state networks - RSNs) and form the anterior
IP
part of the default mode network (DMN) (Damoiseaux et al., 2008). Diminished
R
DMN activity may be attributed to an overall grey matter (GM) reduction, which has
SC
been mostly observed in the frontal and the parietal cortex (Koch et al., 2010). Studies
have also shown that degeneration of cortico-cortical projections, in both healthy and
NU
pathological aging, influences coherent firing of neuronal circuits, thereby affecting
MA
their synchronization (Koenig et al., 2005). Although there is no a unique disturbance
pattern regarding the brain’s co-operative activity, most studies report a reduction of
D
alpha band frequency spectrum (Koenig et al., 2005). Several studies have also shown
P
(Koenig et al., 2005; Park et al., 2008) by means of the Global Field Synchronization
AC
(GFS) and the Synchronization Likelihood (SL) techniques (Stam et al., 2003).
Furthermore, it has been suggested that the loss of gamma band synchronization
among distant brain regions (Koenig et al., 2005). All together these findings may
account for the functional disconnection among distant brain regions that
aging. For instance, studies have reported a synchronization increase in the delta band
latter finding has been attributed to compensatory mechanisms observed within the
theta frequency bands, and linked with poor memory performance (Sankari et al.,
T
2011). Overall, these findings suggest that EEG synchronization could be a valid and
IP
valuable biological marker of early neurodegeneration signs.
R
In the last decades there is a growing interest at investigating whether the
SC
negative effects of aging can be ameliorated, or even delayed, with cognitive or/and
NU
physical training (Colcombe et al., 2006; Boyke et al., 2008; Smith et al., 2009;
Erickson et al., 2011; Simon et al., 2012). At the base of all these cognitive
MA
interventions is the concept of neuroplasticity: the brain is able to adjust and
or creating new ones (Erickson et al., 2011; Mahncke et al., 2006). An example of
TE
such interventions is the IMPACT study, which randomly assigned 487 healthy old
adults to one of two groups: the intervention and the active control group (Smith et al,
P
CE
2009). In the intervention group, participants were intensively trained (40 training
hours in two months) with auditory stimuli by means of computer software named
AC
exercises with varying levels of difficulty, which tap mostly on auditory processing.
Participants in the active control group were also exposed to a computerized task that
did not train specific cognitive processes and received the same hours of “training” as
the cognitive processes directly trained, in the intervention group relative to the active
control group. Most important, there were also significant spillover effects to other
measures that were not directly trained. The authors concluded that the intervention
was effective in improving cognitive performance in old adults. Despite this and other
ACCEPTED MANUSCRIPT
positive findings, there are still some outstanding questions such as finding the most
effective type, and optimal duration of cognitive training, as well as, what the long-
T
Evidence regarding the effects of physical activity on healthy aging appears to
IP
be more robust (Boyke et al., 2008; Billis et al., 2010). There have been several
R
hypotheses put forward to explain the striking positive effects of physical activity on
SC
healthy aging (Colcombe et al., 2006; Erickson et al., 2011). For instance, it has been
NU
reported that aerobic training induces hippocampal neurogenesis (Erickson et al.,
brain perfusion can account for the beneficial effect of physical activity on cognitive
TE
of these approaches and the validation of their effectiveness are currently top research
priorities. For instance, only a few studies (Boyke et al., 2008; Colcombe et al., 2006;
Engvig et al., 2010; Erickson et al., 2011) have actually looked at the functional
thickness of specific regions (Boyke et al., 2008; Colcombe et al., 2006; Engvig et al.,
2010; Erickson et al., 2011). In line with this hypothesis, it has been reported that
intensive verbal memory training with the Method of Loci was associated with an
increase of the cortical thickness in the right fusiform and lateral orbitofrontal cortex
ACCEPTED MANUSCRIPT
among other regions (Engvig et al., 2010). Another study demonstrated that visuo-
motor learning may also induce the plasticity of the mature human brain (Boyke et al.,
2008). Twenty-five healthy old adults were trained for three months in three-ball
T
cascade juggling that, according to the authors, requires dynamic storage and
IP
processing of complex visual stimuli, and visuomotor coordination. The study found
R
that senior citizens who participated in the training group demonstrated a significant
SC
increase in grey matter on the middle temporal area of the visual cortex, on the left
NU
However, the aforementioned effects were not sustained during the 3-month follow-
MA
up period (Boyke et al., 2008).
Studies have also reported brain changes related with physical training.
D
participants, who were trained in aerobic exercises during 6 months. The active
control group was trained during the same period with whole-body stretching and
P
CE
toning exercises. The authors reported an increase in size of the anterior hippocampus
in the intervention group relative to the active control group. These findings were
AC
regarding the groups. It involved 120 healthy older adults who were randomly
assigned to one of two groups described above. The training parameters were three
days per week and the program lasted for one year (Erickson et al., 2011). However, it
was not clear what the functional relevance of these structural changes might be.
distant brain regions. In one related study, Pons et al. (2010) measured structural and
functional alterations associated with aging, using a neural mass model. The authors
T
reported a long-range reduction of synchronization among thalamo-cortical and
IP
cortico-cortical brain regions in healthy and pathological (Mild Cognitive Impairment,
R
MCI) aging. Similarly, Han et al. (2012) reported abnormal functional connectivity
SC
values in patients suffering from MCI as compared with healthy control individuals.
NU
tool for studying interactions among brain regions since it is a non-invasive, relatively
MA
easy to use and cheap technique.
based on the notion of relative wavelet entropy (RWE). The mathematical model
P
CE
adopted employs a simple, accurate, robust and parameter free mathematical notion,
which is combined with a decision making scheme in order to consider both the group
AC
and participant level. To our knowledge, this is the first study that used a
literature findings on both physiological and pathological aging, the current study
mathematical model does not focus on a specific frequency band but on the relative
2. Methods
The participants of the LLM training program performed both cognitive and physical
T
exercise, whereas the ACTIVE protocol was also computerized and had the same
IP
duration / intensity with the intervention protocol (Konstantinidis et al., 2010;
R
Bamidis et al., 2011).
SC
2.1.1 Cognitive Training (CT)
NU
A Greek-adapted version of the Brain Fitness software (© PositScience, USA),
MA
developed during the Long Lasting Memories project (www.longlastingmemories.eu)
(Bamidis et al, 2011), was employed for the cognitive part of the training program
(LLM training). The program consists of six (6) exercises with auditory stimuli
D
TE
(Smith et al., 2009). Each exercise has self-paced levels of difficulty and lasts for 15
minutes. Each individual training session consisted of four (4) out of the six (6)
P
exercises, and lasted for approximately one (1) hour. The whole training protocol
CE
included 1 hour of cognitive training per day for three (3) to five (5) days / per week
AC
during a period of eight (8) weeks. This cognitive program was designed to train
(Mahncke et al., 2006a). Auditory training was selected since the auditory brain
system is better studied than the visual system (Mahncke et al., 2006b) and studies
have shown that age-related cognitive deficits may partially result from reduced
sensory processing in the auditory domain. For instance, it has been shown that high-
frequency auditory loss was associated to impaired speech perception and poor
Physical training was performed through the FitForAll (FFA) platform, which was an
exergaming environment based on the Nintendo Wii Remote and Balance Board
(Billis et al., 2010; Billis et al., 2011). The program consisted of several simple
T
games, carefully designed for senior citizens. These exercises aimed to enhance body
IP
flexibility and strength, as well as, to improve the participants’ physical endurance
R
through aerobic training (Gonzalez-Palau et al., 2013).
SC
2.2 Active Control (AC) Group
NU
In order to control for several potential confound factors such as willingness to adopt
an active aging profile, computer skills and socialization, the IMPACT study (Smith
MA
et al., 2009) included an active control (AC) group rather than a passive control one.
Therefore, it was decided for the purposes of the current piece of work, to also
D
introduce a control design similar to the LLM training program with respect to the
TE
training parameters (computer use, intensity and duration). So, the AC protocol,
P
included viewing of YouTube documentaries with topics derived from nature, art and
CE
history. After the end of the documentary, participants digitally performed a multiple-
choice questionnaire with questions about the documentaries. The AC control group
AC
did not perform any physical training. The whole protocol was also computerized and
A complete set of neuropsychological testing took place twice: (a) prior (1-15 days) to
the study initiation and (b) 1-15 days after completing the training. Details about the
.
ACCEPTED MANUSCRIPT
2.4 Participants
The present study involved 103 participants (53 in the training group and 50 in the
active control group). As mentioned already, the study was part of the multi-centric
T
trials piloted during the Long Lasting Memories (LLM) project in five (5) EU
IP
countries and aiming to promote independent living of senior participants. The
R
majority of the participants (75.73%) were females. Inclusion criteria were; i) age
SC
over 60, ii) MMSE scores above 18 iii) no history of neurological condition (e.g.,
NU
epilepsy, stroke, etc), severe depression or psychiatric disorder, iv) not being currently
committed to the study throughout its duration. Prior to their enrollment, all
D
participants were informed about the study and signed an informed consent form.
TE
During their participation and until the end of the one-year follow up period,
participants were excluded from any other study and/or training program. The study
P
CE
The two groups were matched on age, male-to-female ratio and general
cognitive status as measured by the MoCA and the MMSE tests (see Table 1).
Table 1: Mean age, sex and cognitive status values regarding the LLM training (LLM) and the Active control (AC)
group
equipped with active electrodes attached on a cap fitted to the scalp (EASYCAP). The
T
device consisted of 57 electrodes recording brain signals, 2 reference electrodes
IP
attached to the earlobes and a ground electrode placed at a left anterior position. Both
R
vertical and horizontal electrooculograms (EOG) as well as electrocardiographic
SC
(ECG) activity were also recorded using bipolar electrodes. Electrode impedances
NU
were kept lower than 2 ΚΩs, while the sampling rate was set at 500 Hz. Participants
were instructed to sit in a comfortable armed chair, to close their eyes and to stay
MA
calm for five minutes.
midline were re-referenced by means of the linked earlobe activity. A high-pass filter
P
with a cut-off frequency at 1 Hz was used for removing low-frequency signals, while
CE
a notch filter centered on the 50 Hz eliminated the industrial noise. Both filters were
digital Butterworth filters of 3rd order provided by the Matlab Signal Processing
AC
Toolbox. The Independent Component Analysis (ICA) algorithm was employed for
the detection of artifactual sources. As artifactual sources were regarded linear drifts,
muscle recordings, high frequency noise, heart modulation or artifacts due to the
participant movements. Then, rejection of short data segments that were contaminated
with noise was done by visual inspection. The aforementioned preprocessing steps
were performed through the EEGLAB graphic user interface (Delorme & Makeig,
2004). Finally, 20 seconds of continuous, high quality, artifact-free data were selected
The analysis employed wavelets which are mathematic oscillatory tools ideal
for time-frequency analysis of non-stationary data (Frantzidis et al., 2010). The initial
T
step of the wavelet analysis was the selection of the appropriate mother wavelet,
IP
which determined the basic shape of the wavelet. Then, both time scaling and
R
translation took place resulting in the entire wavelet family. The scaling procedure
SC
was adopted towards the precise representation of frequency varying components,
NU
while translation captured the time-dependent features. The EEG (20 seconds artifact-
free) continuous data were then divided in windows of 128 milliseconds duration. The
MA
first 150 windows were further analyzed. Then, the Orthogonal Discrete Wavelet
transform (ODWT) at periodization mode was used for computing the wavelet
D
The amplitude of the wavelet coefficients quantified the similarity degree among the
wavelet and the actual signal; while the coefficient’s sign indicated whether the
P
CE
correlation was positive or negative (polarity). The discrete version of the wavelet
information; whereas the orthogonal basis facilitated the perfect reconstruction of the
initial brain data. The decomposition scheme employed recursive low-pass filtering
for extracting the activity of each frequency band with optimal resolution. The above
computations were performed through Matlab functions. The window length was the
minimum one that contained at least one coefficient from each frequency band.
ACCEPTED MANUSCRIPT
T
R IP
SC
NU
MA
D
P TE
Figure 1: Visualization of the proposed synchronization analysis framework: Two brain signals from a pair of
CE
electrodes served as an input (first layer). Then, the Orthogonal Discrete Wavelet Transform was adopted and
a recursive decomposition scheme through low-pass filtering resulted in the extraction of the activity for each
AC
frequency band with optimal time-frequency resolution (second layer). The wavelet coefficients for each band
were evaluated and the energy of each band was computed (third layer). Then the similarity degree of the
energy distributions among each electrode pair was computed through the relative wavelet entropy, thereby
For each time window and for each electrode, the wavelet coefficients were
computed for the five frequency bands according to the decomposition scheme
(Figure 1 – upper panel; right part) through equation (1). Since, there were multiple
coefficients were squared and then summed in order to provide the energy of each
ACCEPTED MANUSCRIPT
frequency band. In this formula, j=1…5 denoted the decomposition level, Ck were the
wavelet coefficients and Ej were the energy values for each frequency band:
K
Ej Ck 2 ,
T
j=1…5 (1)
k 1
IP
Then, the total energy (Etot) of the EEG signal was computed as the
R
summation of the energy values for the five frequency bands:
SC
Etot= (2)
NU
Then, the relative energy for each frequency band was computed by simply
MA
dividing its absolute energy value (Ej) with the entire EEG energy (Etot). The relative
energies represented the total energy contribution of a specific rhythmic activity to the
D
whole EEG energy and were positive, non-zero numbers, whereas their summation
TE
Finally, the similarity between the energy distributions among electrodes was
CE
as follows: each electrode activity was compared with each other using electrode
pairs. So, in each time window, there existed N × N electrode comparisons, where N
denoted the number of electrodes (N=57 in this case). The comparison was performed
using the notion of relative wavelet entropy (RWE), whereas pj and qj were the energy
1 – fourth layer) and were repeated for each time window. Then, average values of the
ACCEPTED MANUSCRIPT
matrix were computed for each participant. The larger the value (red color) of a
matrix element, the less synchronization among the corresponding electrode pair; a
zero value (blue diagonal line) was obtained when comparing an electrode with itself.
T
The present study employed the family of bi-orthogonal wavelets of fifth order since
IP
this was a suitable choice when dealing with EEG/ERP data as indicated by Frantzidis
R
et al. (2010).
SC
2.7 Neurophysiological Evaluation of the LLM Intervention
NU
The aforementioned analysis was performed twice for each participant: prior
to the training and after the training. To express synchronization increases as a result
MA
of the intervention, the synchronization matrix of each participant at post condition
Since the total number of likely synchronization features was extremely large
CE
(57 * 57 = 3249 electrode pairs), a feature selection procedure was essential to take
place prior to any statistical analysis. To this extent, a two-fold procedure was
AC
followed. The initial step involved a pre-selection of the electrode pairs in accordance
synchronization) (Babiloni et al., 2004; Babiloni et al., 2009), associated with markers
increase was defined. The threshold value was set arbitrarily at a particularly high
value in order to avoid false positives. More specifically, the average synchronization
ACCEPTED MANUSCRIPT
matrices were computed for both the LLM (N1=53) and the ACTIVE control group
(N2=50) and for each experimental phase (baseline, post-training). Then, a threshold
T
matrices. This process essentially selected only 12 electrode pairs (out of the possible
IP
(57 × 57)-57 = 3192) corresponding to less than 0.4 % of the available (likely)
R
features. So, this resulted in the following electrode pairs which were then further
SC
analyzed:
NU
Table 2: List of selected features used for the neurophysiological evaluation of the LLM training
Mahalanobis distance (Md) based classifier, which was simple, robust and offered a
T
classifier received as an input vector μ which denoted the mean values (centroids) of
IP
the neurophysiological features estimated for the training group (μtr) and the control
R
group (μctrl). The neurophysiological data of each individual participant form a
SC
feature vector χ that was compared with the centroids of the two groups according to
the training or for the control group. So, the above formula computed the distance of
D
TE
each individual feature vector from the mean value of the training or the control data.
3. Results
minus mean synchronizationpost-intervention) for each electrode pair was computed for
versus AC). Results showed that the mean synchronization increase was significantly
larger for the participants of the LLM intervention compared to the AC [F (1, 107) =
3.566, p < .0005, Wilk's λ = 0.308, η2 = .308] (see Table 3). Further one-way
ACCEPTED MANUSCRIPT
ANOVAs for each electrode pair using the Bonferroni correction for multiple
T
IP
Table 3: Means, standard deviations and statistical significance of the candidate features (electrode pairs) that
differed as a function of intervention. The first four rows of Table 3 depict the directed electrode pairs that
R
reached statistical significance after the Bonferroni correction (p-value < 0.004), while the remaining ones
contain the electrode pairs with 0.05≥p-value≥0.004.
SC
Electrode pairs Mean Values p-value
NU
LLM Active
T
R IP
SC
NU
MA
D
P TE
CE
Figure 2: Visualization of the average values for each statistically significant feature, computed for both the
LLM (blue bar) and the Active Control (AC) training group (red bar).
AC
over a user-defined threshold. Then, these pairs were statistically analyzed. However,
such a procedure may increase the risk of false positive results. Aiming to investigate
whether the LLM training resulted in more generalized synchronization increases the
electrode pairs. Therefore, the following areas were defined (see Fig 3).
ACCEPTED MANUSCRIPT
Right Hemisphere Region –RG1 (FC2, Af4, Fp2, F2, F4, F6, F8, C2, C4, C6,
Left Hemisphere Region – RG2 (CP3, CP1, CP5, TP7, P7, P5, P3, P1, PO3,
T
O1, C1, C3, C5, T7)
IP
Right Anterior & Anterior Midline Area - RG3 (Fpz, Afz, Fz, Fp2, Af4, F2,
R
F4, F6, F8)
SC
Right Frontocentral & Right Parietal Region – RG4 (FC2, FC4, FC6, FT8, C2,
C4, C6, T8, CP2, CP4, CP6, TP8, P2, P4, P6, P8)
NU
MA
D
P TE
CE
AC
Figure 3 Definition of regions of interest by extending the electrode pairs identified previously. The left part
defines a right anterior and a left posterior region in order to investigate inter-region synchronization increases
at both directions. The right part defines two regions (another right anterior and a right centroparietal one) for
investigating intra-region synchronization increases.
formation of ROIs was based on a priori information (for a justification see the
randomization technique for selecting the electrode pairs was followed. More
specifically, the available electrode pairs for each ROI were input into a random
number generator which in turn outputs choices of electrode pairs to be used for
ACCEPTED MANUSCRIPT
experiment. So, a fixed number of electrode pairs were randomly selected for each
participant and for each experimental phase (pre & post condition), in accordance
T
with usual bootstrap technique practice. Then, the synchronization analysis
IP
framework was applied for each selected pair. The synchronization among two ROIs
R
was computed as the average value among all the selected electrode pairs. Therefore,
SC
we avoided multiple testing of electrode pairs and estimated a single mean
NU
in the following four cases: two inter-region estimations (RG1-RG2, RG2-RG1) and
MA
two intra-region synchronization computations (RG3 and RG4), thereby resulting in
were averaged, and a single, generalized synchronization value was computed for
TE
each participant and for each phase. This approach was adopted in order to avoid
multiple computations over single electrode pairs that strictly define an interaction
P
usually affected by volume conduction and poor spatial resolution. Thus, instead of
CE
computing synchronization of electrode pairs, this was done for specific ROIs. These
AC
ROIs covered a much broader area (an entire anterior/posterior region of a specific
while there was no need for accurate spatial resolution. The ROI synchronization
synchronization among entire regions and reduced the number of false positives. The
T
R IP
SC
NU
Figure 4 Flowchart demonstration of the computational framework proposed for estimating the
synchronization degree among Regions of Interest (ROIs). Selection of ROIs was performed using a priori
MA
knowledge and already existing aging models. Electrode pairs were randomly selected (bootstrap technique)
and the synchronization degree among two ROIs is estimated (average synchronization among randomly
selected electrode pairs). An overall synchronization value for each participant and for each experimental
phase was estimated for the ROIs under investigation.
D
The mean synchronization increase (pre-post) of the LLM group was .02836 ±
TE
.07791 (standard mean error = .01055); whereas the mean synchronization increase
for the ACTIVE control group was -.02061 ± .01055 (standard mean error = .01070).
P
indeed observed in the majority of the participants in the LLM group, we estimated
the mean value of the synchronization increase/decrease regarding the four (4)
electrode pairs for both the LLM (Figure 2; blue bar) and for the ACTIVE participants
(Figure 2; red bar). Then, we computed the Mahalanobis distance of each feature
Then, each feature instance was classified by comparing its distance from both
T
the training group demonstrated synchronization increases that were closest to the
IP
mean neurophysiological pattern of the LLM group. Regarding the AC group, the
R
classification performance reached 72%, whereas the overall performance was
SC
76.57%.
NU
4. Discussion
In this piece of work, a novel approach was employed to investigate brain changes
MA
associated with a combined cognitive/physical intervention in a group of old adults.
framework based on the notion of relative wavelet entropy (RWE). In agreement with
electrode pairs (FC2-CP3, CP3-FC2, Afz-Fp2, and FC6-P4) and the corresponding
CE
ROIs in the LLM group relative to an active control group. Using a data-mining
AC
technique we found that the majority of the participants (81%) in the LLM group
showed this pattern of improvement. Thus, we can conclude that this is a robust
finding.
quantifying the benefits of cognitive and physical training for promoting healthy
aging. Given that intervention programs are performed in day care centers and small
clinics, there is need for an examination that would be relatively low-cost, non-
invasive, easily performed and with no risk. Although the EEG is a suitable technique
ACCEPTED MANUSCRIPT
since it fulfills all the above criteria, it is rarely used in clinical practice. One of the
main reasons for the relatively low acceptance among clinicians was the vulnerability
T
the extraction of reliable neurophysiological features. Nowadays, these
IP
methodological limitations have been greatly overcome by technological advances
R
allowing simultaneous recordings from dense electrode arrays and facilitating the
SC
acquisition of high quality data through the usage of active electrodes. Artifact
rejection techniques (ICA algorithm) and modern filtering ways (discrete wavelet
NU
decomposition) when combined with elaborate algorithms result in precise and robust
MA
neurophysiological markers that provide insights about the degree of cognitive
deficits.
D
Furthermore, studies suggest that measuring brain activity at rest, and not in
TE
relation to an event, may be more useful when studying aging populations. The
among motor-related brain regions was preserved in resting state condition (Biswal et
al., 1995). This finding indicates that intrinsic activity (resting state) may reflect a
AC
was employed as a powerful tool for linking brain function with the underlying
Structural brain networks, being deterministic in nature, facilitate the occurrence and
completely random signal alterations, but form the brain’s intrinsic dynamic
evolution of functional resting state networks for cognitive function has been
ACCEPTED MANUSCRIPT
T
functioning, and it is also negatively correlated with age (that is the elderly people
IP
show decreased variability). Therefore, we propose that synchronization at rest may
R
be a useful and valid index of brain and cognitive function in old adults, and may,
SC
therefore, be a good outcome measure for assessing the effect of interventions.
NU
On the backbone of the neurophysiological evaluation framework employed
herein was relative wavelet entropy, used as synchronization metric. RWE was
MA
derived as a robust way of quantifying the dynamic cooperation among short EEG
signals (Rosso et al., 2001). This time-evolving entropy notion is based on the
D
orthogonal discrete wavelet transform (ODWT) and was used for its attractive
TE
time-series data of each electrode pair. This makes it an ideal methodology for
biological data since it does not require the unrealistic assumption of time-series
AC
that the signal’s mean value and variance remain constant over time. This limitation
was overcome thanks to the localization characteristics of the wavelets. The adoption
of the ODWT facilitates the optimal time-frequency resolution for each frequency
resulted in four electrode pairs. Two of them were defined by a right frontocentral
co-operation. The third pair involved the electrode placed on an anteriofrontal midline
position (Afz) and the right prefrontal electrode (Fp2), whereas the latter pair
T
single electrode-pairs faces a statistical limitation of producing a high number of false
IP
positives. A partial solution to that problem may be the definition of an extremely
R
high threshold of synchronization increase (greater than 10% in comparison with the
SC
baseline recordings). Even in that case, analysis of single EEG electrodes cannot be
directly linked to specific cortical areas due to the poor spatial capacity of EEG
NU
analysis and the problem of volume conduction. Therefore, findings of
MA
synchronization increases on a few and specific electrode pairs would be of limited
defining ROIs. These ROIs were located on the entire right anteriocentral (RG1), left
centroposterior (RG2), right anterior (RG3) and right posterior (RG4) areas. The
P
Peelle et al., 2013), the age-related decline of the right hemisphere proposed by the
Right hemi-aging model (Albert & Moss, 1988; Dolcos et al., 2002) and the anterior-
Delbeuck et al., 2003; Babiloni et al., 2009). According to the Right hemi-aging
model, the right-hemisphere and especially its posterior part is more vulnerable to
age-related decline in comparison with the left hemisphere (Albert & Moss, 1988).
More recent studies demonstrated structural changes due to aging, which were more
prominent on the prefrontal cortex (Salat et al., 2004). The current piece of work
synchronization through the notion of RWE was employed as a metric. The study
results imply that LLM training may strengthen these vulnerable regions. Previous
T
research efforts attempting to shed light into the neurodegenerative phenomena have
IP
demonstrated the functional disorganization of the mature brain (Delbeuck et al.,
R
2003). These phenomena affect mainly long-distant areas (Babiloni et al., 2004;
SC
Babiloni et al., 2009). Aiming to also investigate whether LLM training would be
effective in senior citizens at high risk (preclinical phase) or already suffering from
NU
dementia, we computed directional synchronization among a right anterior (RG1) and
MA
a left posterior (RG2) region. The aforementioned results demonstrated statistically
significant synchronization increases among these regions for the LLM participants
D
The time duration of the proposed analysis may seem a bit short for analyzing
functional patterns obtained from a large number of electrode pairs. However, it has
P
CE
been previously demonstrated that twenty seconds of data are sufficient for resting-
state analysis with eyes closed in the elderly, while longer epochs did not provide
AC
2007; Hsu et al., 2012). Moreover, the proposed methodology is particularly suitable
for short data segments since the existence of at least one wavelet coefficient for each
frequency band adds to the model’s stability. Apart from its robustness, the proposed
functioning.
ACCEPTED MANUSCRIPT
favors holistic approaches aiming to measure both the brain and cognitive function of
senior citizens. The present study showed an increase in synchronization, in the LLM
T
relative to the AC group, in a network involving four pair electrodes (FC2 - CP3, CP3
IP
- FC2 and FC6 - P4) and the corresponding ROIs. However, EEG signals may be
R
influenced by volume conduction. Therefore, interpretation of EEG functional
SC
connectivity in the source domain should be avoided, or be supported by
neuroimaging studies (Berger & Posner, 2000; Corbetta & Shulman, 2002; Garavan et
NU
al., 2000; Reuter-Lorenz et al., 2000). The finding of increased synchronization in
MA
prefrontal brain areas after a combined cognitive and physical training (LLM) have
important clinical implications, as these areas seem to be the most susceptible to the
D
aging process (Head et al., 2004; Marshall et al., 2011; Raz et al., 2005; Resnick et al.,
TE
2003). Thus, the present results suggest that the LLM intervention may be effective in
different reference methodologies (see Appendix), fits well with the hypothesis that
lobe connectivity (Damoiseaux et al., 2012; Delbeuck et al., 2003; Sanz-Arigita et al.,
within the parietal cortex (intra-parietal connectivity) (Jacobs et al., 2012). Thus, one
and duration of the intervention, the baseline cognitive level of the sample,
T
distance structural and functional connectivity may be more suitable to be applied on
IP
cognitively healthy and highly-functioning elderly participants (Smith et al., 2009).
R
The goal of this training would be to minimize the possibility of dementia onset. On
SC
the other hand, it may be more realistic for patients already suffering from cognitive
NU
functional compensatory mechanisms in order to counteract brain degeneration.
MA
Therefore, the goal of these training programs would be to delay the onset of the
encompass the aforementioned parameters and define the optimal therapeutic strategy
TE
The combination of the RWE synchronization metric with the Mahalanobis distance
individual level and the overall group. The study proposes a pattern recognition
whether and how each senior citizen benefited from the program by comparing her/his
own pattern of synchronization increases with the mean pattern of the training group.
The employment of the proposed approach revealed that 81% of the participants in
ACCEPTED MANUSCRIPT
the intervention group showed similar improvements after training to the overall
group improvement. This suggests that our findings are rather robust. Still 10 out of
T
cognitive/physical intervention. As one cannot be certain about the factors that may
IP
account for these individual differences, further studies are needed to investigate what
R
specific factors (e.g., personality, education, gender, intervention duration and dose
SC
etc) may modulate the effect of such interventions.
NU
4.3 Conclusions
To sum up, the current piece of work presents a novel approach to evaluate brain
MA
functional outcomes of aging-related interventions in general and the LLM training in
that the LLM intervention (combined cognitive and physical training) was effective in
TE
support of our initial hypothesis, we specifically found that synchronization across the
CE
two hemispheres was increased for the LLM group after the intervention. Future
research may address the clinical relevance of this finding, since disturbance of long
AC
range connections among distant brain regions (anterior-posterior) is one of the main
characteristics of pathological aging. It is believed that the present work may trigger
further research aiming at the production of novel quantifiable ways towards the
al, 2014).
approaches
ACCEPTED MANUSCRIPT
To ensure that our main finding was not an artifact of the reference method employed,
ACTIVE controls) by employing (i) common average, (ii) linked earlobes reference
T
(A1+A2)/2 and (iii) the reference already used in our study.
IP
We performed independent samples t-test for each reference methodology and for
R
each electrode pair as suggested by one reviewer. The results are demonstrated in the
SC
following table:
NU
Table A1: Demonstration of the statistical significance of the four electrode pairs for each re-reference
approach
MA
Electrode Study Approach Common Average Linked Earlobes
Pairs
t value p value t value p value t value p value
distant electrode pairs (FC2-CP3 & CP3-FC2) yielded the most significant results
Acknowledgements
T
(www.longlastingmemories.eu).
R IP
The authors would like to thank: A. Semertzidou, M. Karagianni, S. Fasnaki for
SC
neurophysiological data acquisition, A. Billis for technical assistance and software
NU
and V. Zilidou and E. Romanopoulou for pilot executions with seniors.
MA
References
Babiloni, C., Ferri, R., Moretti, R. V., Strambi, A., Binetti, G., Dal Forno, G., Ferreri,
TE
F., Lanuzza, B., Bonato, C., Nobili, F., Rodriguez, G., Salinari, S., Passero, S.,
Rocchi, R., Stam, C. J., Rossini, P. M., 2004. Abnormal fronto-parietal coupling of
brain rhythms in mild Alzheimer's disease: a multicentric EEG study. Eur J Neurosci,
P
19(9), 2583-2590.
CE
Babiloni, C., Ferri, R., Binetti, G., Vecchio, F., Frisoni, G. B., Lanuzza, B., Miniussi,
C., Nobili, F., Rodriguez, G., Rundo, F., Cassarino, A., Infarinato, F., Cassetta, E.,
Salinari, S., Eusebi, F., Rossini, P. M., 2009. Directionality of EEG synchronization
AC
Bamidis, P.D., Konstantinidis, E.I., Billis, A., Frantzidis, C., Tsolaki, M., Hlauschek,
W., Kyriacou, E., Neofytou, M., Pattichis, C.S. 2011. A Web services-based
exergaming platform for senior citizens: The long lasting memories project approach
to e-health care. Proceedings of the Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, EMBS, art. no. 6090694, 2505-2509.
Bamidis et al, Long Lasting Memories Project Deliverable D1.4 Final report,
http://www.longlastingmemories.eu/sites/default/files/LLM_D1.4_final_report_public
_v2.2doc.pdf, July 2012.
Bamidis, P.D.; Vivas; A.B., Styliadis; C., Ladas; A., Frantzidis; C., Klados; M.,
Schlee; W., Papageorgiou, S.G. 2014. A Review of Physical and Cognitive
Interventions in Aging. Neurosci Biobehav Rev. Under Review.
T
IP
Billis, A., Konstantinidis, E., Mouzakidis, C., Tsolaki, M., Pappas, C., Bamidis, P.
2010. A gamelike interface for training seniors’ dynamic balance and coordination.
R
XII Mediterranean Conference on Medical and Biological Engineering and
Computing. Springer, ISBN: 978-3-642-13038-0, 691-694.
SC
Billis, A. S., Konstantinidis, E. I., Ladas, A. I., Tsolaki, M. N., Pappas, C., Bamidis,
P. D. 2011. Evaluating affective usability experiences of an exergaming platform for
NU
seniors. 10th International Workshop on Biomedical Engineering. IEEE, ISBN: 978-1-
4577-0553-3.
MA
Biswal, B., Zerrin Yetkin, F., Haughton, V. M., Hyde, J. S. 1995. Functional
connectivity in the motor cortex of resting human brain using echo‐planar mri., Magn
Reson Med 34(4), 537-541.
D
Blanco, S., Quiroga, R. Q., Rosso, O. A., Kochen, S. 1995. Time-frequency analysis
of electroencephalogram series. Phys Rev E, 51(3), 2624-2631.
TE
Boyke, J., Driemeyer, J., Gaser, C., Buchel, C. May, A., 2008. Training-Induced
Brain Structure Changes in the Elderly. J Neurosci 28(28), 7031-7035.
P
Bullmore, E., Sporns, O. 2009. Complex brain networks: graph theoretical analysis of
CE
Colcombe, S. J., Erickson, K. I., Scalf, P. E., Kim. J. S., Prakash, R., McAuley, E.,
AC
Elavsky, S., Marquez, D. X., Hu, L., Kramer, A. F., 2006. Aerobic Exercise Training
Increases Brain Volume in Aging Humans. J Gerontol A Biol Sci Med Sci 61(11),
1166-1170.
Cotman, C. W., Berchtold, N. C., Christie, L. - A., 2007. Exercise builds brain health:
key roles of growth factor cascades and inflammation. Trends Neurosci 30(9), 464-
472.
Daffner, K. R., Scinto, L. F. M., Weitzman, A. M., Faust, R., Rentz, D. M., Budson,
A. E., & Holcomb, P. J., 2003. Frontal and parietal components of a cerebral network
mediating voluntary attention to novel events. J Cogn Neurosci, 15(2), 294-313.
ACCEPTED MANUSCRIPT
Damoiseaux, J. S., Beckman, C. F., Sanz-Arigita, E. J., Barkhof, F., Scheltens, Ph.,
Stam, C. J., Smith, S. M., Rombouts, S. A. R. B., 2008. Reduced resting-state brain
activity in the “default network” in normal aging. Cereb Cortex 18 (8), 1856-1864.
T
33(4), 828-e19.
IP
Deco, G., Corbetta, M. 2011. The dynamical balance of the brain at rest.
R
Neuroscientist 17(1), 107-123.
SC
Delbeuck, X., Van der Linden, M., Collette, F., 2003. Alzheimer’s Disease as a
Disconnection Syndrome?. Neuropsychol Rev 13 (2), 79-92.
NU
Delorme, A., Makeig, S. 2004. EEGLAB: an open source toolbox for analysis of
single-trial EEG dynamics including independent component analysis. J Neurosci
Methods 134(1), 9-21.
MA
Dolcos, F., Rice, H. J., Cabeza, R., 2002. Hemispheric asymmetry and aging: right
hemisphere decline or asymmetry reduction. Neurosci Biobehav Rev 26(7), 819-825.
Engvig, A., Fjell, A. M., Westlye, L. T., Moberget, T., Sundseth, O., Larsen, V. A.,
D
Walhovd, K. B., 2010. Effects of memory training on cortical thickness in the elderly.
TE
Erickson, K. I., Voss, M. W., Prakash, R. S., Basak, C., Szabo, A., Chaddock, L.,
P
Kim, J. S., Heo, S., Alves, H., White, S. M., Wojcicki, T. R., Mailey, E., Vieira, V. J.,
Martin, S. A., Pence, B. D., Woods, J. A., McAuley, E., Kramer, A. F., 2011.
CE
Exercise training increases size of hippocampus and improves memory. Proc Natl
Acad Sci USA 108(7), 3017-3022.
AC
Ewers, M., Sperling, R. A., Klunk, W. E., Weiner, M. W., Hampel, H., 2011.
Neuroimaging markers for the prediction and early diagnosis of Alzheimer's disease
dementia. Trends Neurosci 34 (8), 430-442.
Frantzidis, C. A., Bratsas, C., Papadelis, C. L., Konstantinidis, E., Pappas, C., &
Bamidis, P. D. 2010. Toward emotion aware computing: an integrated approach using
multichannel neurophysiological recordings and affective visual stimuli. IEEE Trans
Inf Technol Biomed 14(3), 589-597.
Garavan, H., Ross, T. J., Li, S. J., Stein, E. A. 2000. A parametric manipulation of
central executive functioning. Cereb Cortex, 10(6), 585-592.
Garrett, D. D., Kovacevic, N., McIntosh, A. R., & Grady, C. L. 2010. Blood oxygen
level-dependent signal variability is more than just noise. J Neurosci 30(14), 4914-
4921.
ACCEPTED MANUSCRIPT
Gonzalez-Palau, F., Franco, M., Toribio, J. M., Losada, R., Parra, E., Bamidis, P.
2013. Designing a Computer-based Rehabilitation Solution for Older Adults: The
Importance of Testing Usability, PsychNol 11(2), 119-136.
Gudmundsson, S., Runarsson, T. P., Sigurdsson, S., Eiriksdottir, G., Johnsen, K.,
2007. Reliability of quantitative EEG features. Clin Neurophysiol, 118(10), 2162-
T
2171.
IP
Han, S. D., Arfanakis, K., Fleischman, D. A., Leurgans, S. E., Tuminello, E. R.,
R
Edmonds, E. C., Bennett, D. A., 2012. Functional Connectivity Variations in Mild
Cognitive Impairment: Associations with Cognitive Function. J Int Neuropsychol Soc
SC
18(1), 39-48.
Head, D., Buckner, R. L., Shimony, J. S., Williams, L. E., Akbudak, E., Conturo, T.
NU
E., McAvoy, Morris, J. C., Snyder, A. Z. 2004. Differential vulnerability of anterior
white matter in nondemented aging with minimal acceleration in dementia of the
Alzheimer type: evidence from diffusion tensor imaging. Cereb Cortex, 14(4), 410-
MA
423.
Hsu, Y. F., Huang, Y. Z., Lin, Y. Y., Tang, C. W., Liao, K. K., Lee, P. L., Tsai, Y. A.,
Cheng, H. L., Cheng, H., Chem, C., M., Lee, I. H., 2012. Intermittent theta burst
D
Jacobs, H. I., Van Boxtel, M. P., Jolles, J., Verhey, F. R., & Uylings, H. 2012. Parietal
cortex matters in Alzheimer's disease: an overview of structural, functional and
P
Knyazeva, M. G., Carmeli, C., Khadivi, A., Ghika, J., Meuli, R., Frackowiack, R. S.,
2013. Evolution of source EEG synchronization in early Alzheimer's disease.
AC
Koch, W., Teipel, S., Mueller, S., Buerger, K., Bokde, A. L. W., Hampel, H., Coates,
U., Reiser, M., Meindl, T., 2010. Effects of aging on default mode network activity in
resting state fMRI: Does the method of analysis matter? Neuroimage 51 (1), 280-287.
Konstantinidis. E. I., Billis, A., Hlauschek, W., Panek, P., Bamidis, P. D., 2010.
Integration of cognitive and physical training in a smart home environment for the
elderly people. Studies in Health Technology and Informatics 160 (Pt. 1), 58-62.
Koenig, T., Prichep, L., Dierks, T., Hubl, D., Wahlund, L. O., Hohn, E. R., Jelic, V.,
2005. Decreased EEG synchronization in Alzheimer’s disease and mild cognitive
impairment. Neurobiol Aging 26 (2), 165-171.
Lazarov, O., Robinson, J., Tang, Y.-P., Hairston, I. S., Korade-Mirnics, Z., Lee, V.
M.-Y., Hersh, L. B., Sapolsky, R. M., Mirnics, K., Sisodia, S. S., 2005.
ACCEPTED MANUSCRIPT
Mahncke, H. W., Connor, B. B., Appelman, J., Ahsanuddin, O. N., Hardy, J. L.,
Wood, R. A., Joyce, N. M., Boniske, T., Atkins, S. M., Merzenich, M. M., 2006a.
Memory enhancement in healthy older adults using a brain plasticity-based training
T
program: A randomized, controlled study. Proc Natl Acad Sci U S A 103(33), 12523-
IP
12528.
R
Mahncke, H. W., Bronstone, A., Merzenich, M. M., 2006b. Brain plasticity and
functional losses in the aged: scientific bases for a novel intervention. Prog Brain
SC
Res, 157, 81-109.
Marshall, G. A., Rentz, D. M., Frey, M. T., Locascio, J. J., Johnson, K. A., Sperling,
NU
R. A. 2011. Executive function and instrumental activities of daily living in mild
cognitive impairment and Alzheimer's disease. Alzheimers Dement 7(3), 300-308.
MA
McIntosh, A. R., Korostil, M. 2008. Interpretation of neuroimaging data based on
network concepts. Brain Imaging Behav 2(4), 264-269.
Obisesan, T. O., Gillum, R. F., Johnson, S., Umar, N., Williams, D., Bond, V.,
Kwagyan, J., 2012. Neuroprotection and Neurodegeneration in Alzheimer’s Disease:
D
Role of Cardiovascular Disease Risk Factors, Implications for Dementia Rates, and
TE
Prevention with Aerobic Exercise in African Americans. Int J Alzheimers Dis 2012
(2012), Article ID 568382, 14 pages.
P
Park, Y. – M., Che, H. – J., Im, C. – H., Jung, H. – T., Bae, S. – M., Lee, S. – H.,
CE
2008. Decreased EEG synchronization and its correlation with symptom severity in
Alzheimer's disease. Neurosci Res 62 (2), 112-117.
Peelle, J. E., Chandrasekaran, K., Powers, J., Smith, E. E., Grossman, M., 2013. Age-
AC
Pons, A. J., Cantero, J. L., Atienza, M., Garcia-Ojalvo, J., 2010. Relating structural
and functional anomalous connectivity in the aging brain via neural mass modeling.
Neuroimage 52(3), 848-861.
Raz, N., Lindenberger, U., Rodrigue, K. M., Kennedy, K. M., Head, D., Williamson,
A., Dahle, C., Gerstof, D., Acker, J. D., 2005. Regional Brain Changes in Aging
Healthy Adults: General Trends, Individual Differences and Modifiers. Cereb Cortex
15(11), 1676-1689.
Resnick, S. M., Pham, D. L., Kraut, M. A., Zonderman, A. B., & Davatzikos, C. 2003.
Longitudinal magnetic resonance imaging studies of older adults: a shrinking brain. J
Neurosci 23(8), 3295-3301.
ACCEPTED MANUSCRIPT
Reuter-Lorenz, P. A., Jonides, J., Smith, E. E., Hartley, A., Miller, A., Marshuetz, C.,
& Koeppe, R. A. 2000. Age differences in the frontal lateralization of verbal and
spatial working memory revealed by PET. J Cogn Neurosci, 12(1), 174-187.
Rosso, O. A., Blanco, S., Yordanova, J., Kolev, V., Figliola, A., Schurmann, M., &
Basar, E. 2001. Wavelet entropy: a new tool for analysis of short duration brain
T
electrical signals. J Neurosci Methods 105(1), 65-76.
IP
Salat, D. H., Buckner, R. L., Snyder, A. Z., Greve, D. N., Desikan, R. S, R., Busa, E.,
R
Morris, J. C., Dale, A. M., Fischl, B. 2004. Thinning of the Cerebral Cortex in Aging.
Cereb Cortex 14 (7), 721-730.
SC
Sankari, Z., Adeli, H., Adeli, A., 2011. Intrahemispheric, interhemispheric, and distal
EEG coherence in Alzheimer’s disease. Clin Neurophysiol 122 (5), 897-906.
NU
Sanz-Arigita, E. J., Schoonheim, M. M., Damoiseaux, J. S., Rombouts, S. A. R. B.,
Maris, E., Barkhof, F., Scheltens, Ph., Stam, C. J., 2010. Loss of ‘Small-World’
MA
Networks in Alzheimer's Disease: Graph Analysis of fMRI Resting-State Functional
Connectivity. PLoS One 5 (11), e13788.
Smith, G. E., Housen, P., Yaffe, K., Ruff, R., Kennison, R. F., Mahncke, H. W.,
Zelinski, E. M., 2009. A Cognitive Training Program Based on Principles of Brain
P
Stam, C. J., Van Der Made, Y., Pijnenburg, Y. A. L., Scheltens, Ph., 2003. EEG
synchronization in mild cognitive impairment and Alzheimer's disease. Acta Neurol
AC
Zhang, Z., Liu, Y., Jiang, T. Zhou, B., An, N., Dai, H., Wang, P., Zhang, X., 2012.
Altered spontaneous activity in Alzheimer's disease and mild cognitive impairment
revealed by Regional Homogeneity. Neuroimage 59(2), 1429-1440.
Highlights
T
Robust, cheap and easily applicable to day care centers
IP
Results were in line with existing neuroscience findings
R
SC
NU
MA
D
P TE
CE
AC