525

15

EEG Analysis of Neurodegenerative Diseases

15.1 Introduction

Neurodegenerative diseases are a heterogeneous group of disorders that are characterized

by the progressive degeneration of the structure and function of the central nervous system
or peripheral nervous system (CNS). It is a collective term used to describe various symp-
toms of cognitive decline, such as forgetfulness and a symptom of several underlying dis-
eases and brain disorders. This disease umbrella includes major clinicopathological entities
such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and frontotemporal dementia
(FTD), as well as less common disease subtypes such as progressive supranuclear palsy
(PSP), corticobasal degeneration (CBD), and multiple systems atrophy [1].
Neurodegenerative diseases are challenging to diagnose because there is marked heter-
ogeneity in clinical and pathological biomarkers, some of which are only identifiable by
postmortem examination and not under any in vivo assessment [2]. These degenerative dis-
eases are associated with neuronal and synapsis loss, neuroinflammation, gliosis and vas-
cular abnormalities in specific regions of the brain, and with the deposition of certain
proteins displaying toxic effects; amyloid β and τ in AD; α-synuclein in PD, Lewy body
dementia (LBD) and multiple system atrophy (MSA); τ in PSP and CBD; and transactive
response DNA-binding protein 43 kDa (TDP-43) in FTD and amyotrophic lateral sclerosis
(ALS). Unfortunately, on neuropathological examination most patients show two, three, or
even four of such proteopathy features which makes the diagnosis of these diseases very
challenging.
By applying machine learning to semiquantitative pathological features in a large dataset
of autopsy cases, Cornblath et al. have reported that six transdiagnostic clusters of neuro-
degenerative disease summarize relevant aspects of their phenotypes, and that the pheno-
types can be reasonably predicted from cognitive scores, biomarkers in cerebrospinal fluid
(CSF) or genotype [1]. In their classification they used nine pathological features including
amyloid β and τ levels, α-synuclein, TDP-43, ubiquitin, neuritic plaques, angiopathy, gliosis,
and neuron loss as the features.
Dementia causes dysfunction relative to a person’s previous level of social and occupa-
tional functioning. This disease is a very common medical condition that can result from
diverse causes. One new case of dementia is registered every four seconds in the world.
Dementias are classified based on clinical, genetic, and neuropathological features. The
most common cause of dementia is AD. Other major classes of disorders that cause
EEG Signal Processing and Machine Learning, Second Edition. Saeid Sanei and Jonathon A. Chambers.
© 2022 John Wiley & Sons Ltd. Published 2022 by John Wiley & Sons Ltd.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
526 15 EEG Analysis of Neurodegenerative Diseases

dementia are vascular dementias, psychiatric diseases, and other neurodegenerative dis-
eases, including FTD and LBD.
Chronic conditions that alter or damage nerve cells and synapses involved in cognition
are the biological basis of dementia. Currently, there is no effective treatment to prevent or
reverse the underlying disease process for the most common neurodegenerative diseases of
ageing that cause dementia. The diagnosis, however, is routine and is often achieved by tra-
ditional memory tests (such as Addenbrooke’s Cognitive Examination [ACE] [3]). The ACE
involves the assessment of attention/orientation, memory, language, verbal fluency, and
visuospatial skills.
Electroencephalogram (EEG)-based tests are becoming clinically applicable by evaluat-
ing the brain connectivity, mostly explained in Chapter 8 of this book and some in the fol-
lowing sections.
Based on some new hypothesis, the methods to rule out other possibly treatable diseases
(such as depression, vitamin B12 deficiency, hydrocephalus, and hypothyroidism) are also
used for detecting neurodegenerative diseases such as AD. These methods examine the
blood protein and other deficiencies and often benefit from machine learning techniques.
Dementia is commonly seen in older people and characterized by a decline in cognitive
performance. There are approximately 47 million people worldwide with dementia and
there are 10 million new cases every year. It is a major cause of disability and dependence
and impacts on the physical, psychologic, and social well-being of families and carers.
Gait and balance impairments are common in people with dementia and significantly
increase the risk of falls, injury, institutionalization, hospitalization, morbidity, and death
after a fall particularly in elderlies.
Dementia may be in any of the following four stages:

• Stage 1: Mild cognitive impairment (MCI): characterized by general forgetfulness. This

affects many people as they age but it only progresses to dementia for some. MCI shares
some features with AD. The cognitive deficits of MCI subjects are limited to memory but
their activities of daily living are preserved.

• Stage 2: Mild dementia: people with mild dementia experience cognitive impairments that
occasionally impact their daily life. Symptoms include memory loss, confusion, person-
ality changes, getting lost, and difficulty in planning and performing their daily tasks.

• Stage 3: Moderate dementia: in this stage daily life becomes more challenging, and the
individual may need more help. The symptoms become more severe compared to the
mild case. Individuals may need help getting dressed and combing their hair. They
may also show significant changes in personality; for instance, becoming suspicious or
agitated for no reason. They may face sleep disturbances too.

• Stage 4: Severe dementia: at this stage, the symptoms worsen considerably. There may be a
loss of ability to communicate, and the individual might need full-time care. Simple tasks,
such as sitting and holding one’s head up become impossible and they hardly control their
bladders.
Dementias can be caused by brain cell death or a neurodegenerative disease. Hence, as
well as progressive brain cell death, like that seen in AD, dementia can be caused by a head
injury, a stroke, or a brain tumour, among other causes. The causes include vascular demen-
tia (also called multi-infarct dementia), resulting from brain cell death caused by conditions
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.2 Alzheimer’s Disease 527

such as cerebrovascular disease, for example, stroke. This prevents normal blood flow,
depriving brain cells of oxygen. It may also result from brain injury. Some types of traumatic
brain injury – particularly if repetitive, such as those received by sports players – have been
linked to certain dementias appearing later in life. Although with less probability, a single
brain injury raises the likelihood of having a degenerative dementia such as AD. Dementia
can also be due to prion diseases, for instance, Creutzfeldt–Jakob disease (CJD), human
immunodeficiency virus (HIV) infection in which the virus can damage the brain cells,
and reversible factors, i.e. those dementias that can be treated by reversing the effects of
underlying causes, including medication interactions, depression, vitamin deficiencies,
and thyroid abnormalities.

15.2 Alzheimer’s Disease

In 1901, Karl Deter, a railway worker, admitted his 51-year old wife, Auguste, to a psychi-
atric institution with symptoms of memory loss, confusion, violent outbursts, and inability
to use language. According to Karl, Auguste’s symptoms began to emerge in the late 1890s
and were quite uncommon for the person he had come to know. She had become increas-
ingly fearful and anxious and would sometimes scream loudly for several hours. Her con-
dition became so overwhelming and debilitating that her family could no longer manage
her care. For the next five years until she passed away, Auguste remained at the institution
and was observed by a German psychiatrist, Alois Alzheimer. After her death, Alzheimer
performed histological studies on Auguste Deter’s brain tissue. In the process, he discovered
two abnormalities: large abnormal clumps had formed between neurons, and rope-like tan-
gles had formed inside neurons. Calling these abnormalities ‘a peculiar disease of the cor-
tex’, Alzheimer presented his findings at a 1906 psychiatric conference in Germany,
marking the first documented case of what is now known as AD. Over the next five years,
11 similar cases were reported in medical journals. Today, more than one hundred years
later, there are approximately 35.6 million people suffering from AD worldwide. To the
shock of many, the 2010 World Alzheimer Report projected that this number will almost
double to 65.7 million cases by 2030 and will more than triple to 115.4 million cases by 2050.
AD is categorized as an incurable, degenerative neurological disorder. In popular lan-
guage, AD is commonly used interchangeably with dementia, but they are not the same.
Dementia is the general term used to describe a decline in cognitive abilities (memory,
thinking, language, judgement, and behaviour) and can be associated with a number of dif-
ferent neurological or psychiatric diseases. As an important distinction, AD is the most com-
mon cause of dementia, accounting for 60–80% of all dementia diagnoses. It is primarily
characterized by a loss of short-term memory as well as impairments in other cognitive
functions such as language, problem solving, attention, and orientation. As the disease pro-
gresses, severe mood swings and unprovoked aggression are common. Eventually, the
patient loses almost all language ability and motor function.
For the most definitive diagnosis, scientists use postmortem analysis of brain tissue. The
tissue abnormalities first defined by Alzheimer as clumps and tangles among neurons in the
cortex are more specifically defined as amyloid plaques and neurofibrillary tangles. Plaques
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
528 15 EEG Analysis of Neurodegenerative Diseases

are deposits of a protein called beta-amyloid, which scientists believe is one of the earliest
signs of the disease process. Over time, the build-up of this protein in the brains of Alzhei-
mer’s patients eventually disrupts effective communication between neurons. Scientists
hypothesize that these amyloid accumulations can themselves trigger the formation of
neurofibrillary tangles, which are made of a different protein called p-tau and have a
distinct string-like appearance. Like plaques, these tangles also interfere with neurons,
but they do so by accumulating inside of them in large amounts and eventually killing them.
These tissue pathologies and the behavioural symptoms are not always tightly correlated.
However, plaques and tangles may appear in the brain well before symptoms of dementia
appear which can be detected by magnetic resonance imaging (MRI).
Although AD is commonly diagnosed in people over the age of 65, early-onset cases –
appearing at ages 30–60 and accounting for only 1% of total cases – represent a rare subset
in which great advances have been made. This form of the disease can run in families.
Studies of these families have identified several genes that, when altered, can cause
AD. These genes implicate beta-amyloid as a central player in the disease and formation
of plaques in the brain as an early manifestation of a disease process in the brain. Recent
studies imaging the brains of older adults with and without dementia suggest that these
plaques may first appear as early as 10 years before the onset of clinical symptoms. The
growth of this category may be concurrent with a growing recognition of the disease
among diagnosticians, but it nevertheless causes great concern and contributes to the
expanding effect of the disease.
The cost of care for those with AD can leave supporting families emotionally and finan-
cially debilitated. Currently, there are no cures for this disease, but there are two types of
drugs available that slow disease progression. Unfortunately, both types provide only
marginal improvements in cognitive symptoms. Making matters worse, these drugs tend
to work for only about half the patients who take them, and among those patients, the
effects are moderate and temporary. Based on the discovery that beta-amyloid plaques
appear early on in the disease, even before symptoms appear, many drug companies
are trying to inhibit the production of this protein or target its removal as a possible
strategy to treat the disease.
As life expectancy is on the rise, increasingly more people are likely to develop AD.
Currently, there are more than one billion people over the age of 60. The growing number
of potential sufferers and the inevitable increase in care demands will have an enormous
economic effect on health care and social services.

15.2.1 Application of Brain Connectivity Estimation to AD and MCI

Brain connectivity analyses show considerable promise for understanding how our neural
pathways gradually break down in ageing and AD. Nevertheless, we know very little about
how the brain’s networks change in AD, and which metrics are best to evaluate these
changes. Also, connectivity estimation can be applied to EEG, magnetoencephalogram
(MEG), or functional magnetic resonance images (fMRI) [4].
Following the methods described in Chapter 8, rooted in the Granger causality (GC)
metric [5], magnitude squared coherence – cx,y(f) – is a widespread connectivity metric
applied to EEG and MEG, which provides a measure of the linear dependencies between
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.2 Alzheimer’s Disease 529

two signals x and y as a function of frequency [6, 7]. This complex-valued metric, as defined
in Chapter 8 of this book, is:

S2xy f
cx,y f = = Re cx,y f + iIm cx,y f 15 1
S2xx f S2yy f

where S2xy is cross-power spectrum and S2xx and S2yy are respectively the power spectra of x
and y.
It has been hypothesized that AD decreases cx,y(f) in alpha (α) and beta (β) but there are
no conclusive findings about the delta (δ) and theta (θ) bands [8, 9]. Furthermore, AD
produces opposite changes in features computed from different spectral bands [10, 11].
Thus, it is advisable to analyse different spectral bands separately [12]. Yet, cx,y(f) has
several limitations. Firstly, spurious correlations could appear due to the volume conduc-
tion [11, 13, 14]. This effect is due to the fact that nearby channels are likely to record
activity from identical currents. This leads to abnormally high correlations in the results
that reflect volume conduction rather than actual connectivity. Other alternatives, such as
the imaginary part of the coherency [14, 15] or the phase lag index (PLI) [11], have been
proposed.
The imaginary part of coherency (iCOH), which is Im(cx,y(f)), is used to circumvent the
volume-conduction effects in functional connectivity estimation. An improved functional
connectivity estimation using the iCOH measure in comparison to coherence analysis
has been demonstrated and showed transient interactions between left–right motor cortical
signals as a function of time and frequency in a real dataset. However, due to its exclusive
dependency on the iCOH, the connectivity estimate based on iCOH becomes negligible in
some situations even in the presence of a significant true interaction, e.g. the phase differ-
ence between two signals is near zero or kπ. Later some improvements on this limitation
were achieved by proposing the PLI [11, 16] and the weighted PLI (wPLI) [17], demon-
strated by simulations based on the Kuramoto model as well as with real data. In this latter
work the effect of volume conductivity on each coherence estimation technique has been
analyzed in detail. It has been shown that iCOH cannot be spuriously increased by volume
conduction of independent sources. The PLI and iCOH between the signals from two
sensors are non-equiprobable. Compared to iCOH, PLI has the advantage of being less
influenced by phase delays. However, sensitivity to volume conduction and noise, and
capacity to detect the changes in phase synchronization are hindered by the discontinuity
of the PLI, as small perturbations turn phase lags into leads and vice versa. To solve this
problem, we introduce a related index, namely the wPLI. Differently from PLI, in wPLI
the contribution of the observed phase leads and lags is weighted by the magnitude of
the imaginary component of the cross-spectrum [17].
In [15] the authors proposed the envelope of the imaginary coherence (EIC) as another
measure for functional connectivity and compared it with lagged coherence, iCOH, PLI,
and wPLI, using bivariate autoregressive and stochastic neural mass models.
They provided the definition for modified iCOH as:

EIm X f Y f ∗
iCOHmx,y = 15 2
E H Im X f Y f ∗
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
530 15 EEG Analysis of Neurodegenerative Diseases

where H(∙) refers to Hilbert transform operation. The phase lock value (PLV) is defined as:

j phase of X f − phase of Y f
PLVx,y f = E e 15 3

PLI is a measure of the asymmetry of the distribution of phase differences between two
signals. It reflects the consistency with which one signal is phase leading or lagging with
respect to another signal which is defined as:

PLIx,y f = E sgn phase of X f − phase of Y f 15 4

where sgn(∙) refers to sign. wPLI is defined as:

∗
E Im X f Y f
wPLIx,y f = ∗ 15 5
E Im X f Y f

and, lagged coherence, lCOH is defined as:

Im cX,Y f 2
lCOHx,y f = 2 15 6
1 − Re cX,Y f

EIC is simply

EICx,y = H Im cx,y f 15 7

Also, a modified version of EIC is defined as:

E Im X f Y f ∗
EICmx,y = H 15 8
E H Im X f Y f ∗

Finally, they conclude and claim that EIC and iCOH demonstrate superior results
compared with other functional connectivity estimates.
These measures are usually applied in a channel-wise manner, producing a very high
number of variables [13, 18]. All these studies and experiments clearly show that the
brain connectivity significantly changes for many of these neurodegenerative diseases
particularly AD.
A possible option is to try to estimate the equivalent current dipoles [1, 19, 20]. Once
the currents are estimated, the same connectivity metrics used in the channel-wise
analyses can be applied to them. Moreover, the volume-conduction effect does not affect
these dipole-based connectivity assessments. However, the solution of the inverse
problem is not unique and the choice of the reconstruction algorithm and other a
priori assumptions influence the dipole estimation leading to potentially misleading
results [12, 14].
A research study was carried out by Escudero et al. [21] for differentiation between NC,
AD, and MCI using MEG signals. In their research they pre-processed the signals before
assessing the coherence between the selected brain zones.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.2 Alzheimer’s Disease 531

LA RA
LA RA
LA RA
LL LA RA RL RL
LL LA LA RA RA RL

LL LL LA RA RL
LL LL LC RC RL RL RL
LL LL RL RL
LC LC RC
RC RL
LL RL
LL LL RL RL
LL LL LC LC RC RC RL
RL
LL RL
LL LL LL LC LC RC RC RL RL
LL RL RL
LL
LL LL LL LC LC RC RC RL RL RL RL
LL
LC RL RL
LL LL LC RC RC
LL RL RL RL
LL LL
LC RC RL
LL LP RP
LP RP RL
LL LP RP RL RL
LL LL
LP RP RL
LL LP RP RP RP
LP LP
LP RP RP
LP RP
LP
RP RP
LP LP RP
LP
LP RP
RP
LP
LP RP

Figure 15.1 Distribution of the MEG sensors into left central (LC), anterior (LA), lateral (LL) and
posterior (LP); and right central (RC), anterior (RA), lateral (RL) and posterior (RP) regions. Midline
sensors appear empty. The lines depict the region boundaries.

In this work scalp electrode space has been divided into nine regions as illustrated in
Figure 15.1. Instead of measuring the connectivity directly, MEG signals have been
processed using the empirical mode decomposition (EMD) and blind source separation (BSS).
In the proposed algorithm, the EMD technique was used to decompose each MEG
channel into a set of intrinsic mode functions (IMFs). Then, a constraint BSS block extracts
the representative delta, theta, alpha, and beta rhythms of each region. In addition to the
corresponding spectral coherence, magnitude squared coherence (MSC) is used here to
measure the brain synchrony. It quantifies linear correlations as a function of frequency.
MSC is bounded between 0 and 1. This measure can detect the linear synchronization
between two signals, but it does not discriminate the directionality of the coupling
[8.50]. A simplified block diagram of the algorithm is presented in Figure 15.2. In
Figure 15.3a and b a channel of MEG signal and its corresponding IMFs computed using
the EMD algorithm can be viewed.
To select the right IMFs in terms of centre frequency from each region, the IMFs have
been clustered based on their amplitudes, frequency, and proximity to any one of the four
centre frequencies. As depicted in Figure 15.4 the k-means algorithm has been used to select
a reference signal frequency for each band. This is necessary since the frequency band cen-
tres change from subject to subject.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
532 15 EEG Analysis of Neurodegenerative Diseases

Region i Region j

MEG channel MEG channel MEG channel MEG channel

EMD EMD EMD EMD

Pool of IMFs for region i Pool of IMFs for region j

Selection of references for Selection of references for

region i with k-means region j with k-means

Extraction of Extraction of
rhythms with cBSS rhythms with cBSS

Assessment of c(f) between

region i and j for each band

Figure 15.2 Block diagram of the spectral coherency, c(f), measure.

Quantitative measurement of spectral coherency without or with regional extraction of

rhythms has led to the charts in Figure 15.5 respectively in the left and right bar graphs.
As a result, raw MEG measurements of subject group pairs produced c(f) accuracy rates of
51.6–61.1%, compared to 66.7–77.3% for the pairs of subject groups when rhythms were
extracted.

15.2.2 ERP-Based AD Monitoring

Advanced techniques in signal processing can be employed to detect and reveal the changes
in event-related potentials (ERPs), particularly for P300 subcomponents, due to AD. P3b sub-
component elicits as the result of novel or target stimuli in an odd-ball paradigm experiment.
For AD patients both amplitude and latency of P3a changes compared to the healthy subjects.
For AD the P3b amplitude reduces whereas P3a amplitude does not have any significant
change compared with those of healthy subject. Also, the relative (to P3a) latency for P3b
significantly increases as the result of AD [22]. Figure 15.6 clearly shows the changes in P3b.

15.2.3 Other Approaches to EEG-Based AD Monitoring

The authors of [11] examined changes in the large-scale structure of resting-state brain
networks in patients with AD and compared with non-demented controls, by applying
concepts from graph theory to the MEG signals.
For the main frequency bands, synchronization between all pairs of MEG channels was
assessed using PLI-weighted connectivity networks and characterized by a mean clustering
coefficient and path length.
10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
533
15.2 Alzheimer’s Disease

(a) A 10 second segment of one EMG channel and (b) its corresponding IMFs.
10

10

10

10

10

10

10

10
10

10
9

9
9

9
8

8
8

8
7

7
7

7
6

6
6

Time (s)
Time (s)

(b)
(a)

5
5

4
4

3
3

2
2

1
1

Figure 15.3
0

0
0

0.2

–0.2
0.2

–0.2

0.2

–0.2

0.2

–0.2

0.2

–0.2

0.2

–0.2

0.2

–0.2

0.2

–0.2

0.2

–0.2
0.4

0.2

–0.2

–0.4
Amplitude (pT) Amp. (pT)
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
534 15 EEG Analysis of Neurodegenerative Diseases

10
Frequency
of the IMFs
9
δ centroid
8 θ centroid
α centroid
7 β centroid
Number of elements

0
1 5 10 15 20 25 30
Frequency (Hz)

Figure 15.4 Reference selection using the k-means algorithm.

Using this method, the AD patients showed a decrease of mean PLI in the lower alpha and
beta bands. In the lower alpha band, the clustering coefficient is defined as [11]:

k i l i wik wil wkl

l k
Ci = 15 9
k i l i wik wil
l k

where wij is the weight of the link between vertices i and j and (the average) path length
defined as [11]:

N N −1
1
Lw = wij 15 10
N N −1 i=1 j i

were both decreased in AD patients. The mean values for the above two parameters in
different frequency bands are calculated and used in classification between healthy and
AD subjects [11]. Network changes in the lower alpha band were better explained by a
‘Targeted Attack’ model than by a ‘Random Failure’ model. Thus, the AD patients dis-
play a loss of resting-state functional connectivity in lower alpha and beta bands even
when a measure insensitive to volume-conduction effects, such as PLI, is used. More-
over, the large-scale structure of lower alpha band functional networks in AD is more
random.
10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Control
MCI

Beta
AD

Alpha
Bands
Theta

Spectral coherency levels without (left) and with (right) regional rhythms extraction.
Delta
0.28

0.26

0.24

0.22

0.2

0.18

0.16

0.14

0.12
c(f) for the adaptively extracted activity

Control
MCI

Beta
AD

Alpha
Bands
Theta
Delta

Figure 15.5
0.28

0.26

0.24

0.22

0.2

0.18

0.16

0.14

0.12
c(f) for the unprocessed channels
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
536 15 EEG Analysis of Neurodegenerative Diseases

Averaged over Averaged over Averaged over Averaged over

50–130 s 90–170 s 130–210 s 170–250 s
25 25 25 25
P3a
P3a
20 20 20 20
P3a
P3b P3b
Amplitude (μV)

15 15 15 15 P3a

10 10 10 10

5 5 5 5

0 0 0 0

–5 –5 –5 –5

–10 –10 –10 –10

250 500 250 500 250 500 250 500
Time (ms) Time (ms) Time (ms) Time (ms)
(a)

Averaged over Averaged over Averaged over Averaged over

50–130 s 90–170 s 130–210 s 170–250 s
20 20 20 20

15 P3a 15 15 15
P3a
10 10 10 P3a 10 P3a
Amplitude (μV)

5 5 5 5

0 0 0 0
P3b P3b
P3b P3b
–5 –5 –5 –5

–10 –10 –10 –10

250 500 250 500 250 500 250 500
Time (ms) Time (ms) Time (ms) Time (ms)

Time frame average Overall average

(b)

Figure 15.6 ERPs recorded using the Cz electrode for (a) healthy and (b) AD patient. The bold graphs
are averaged over a smaller number of 20 second consecutive frames (denoted on the top) and the
dashed graphs are the overall average ERPs. There is clear decrease in amplitude and increase in the
relative (to P3a) latency of P3b for AD patients.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.4 Parkinson’s Disease 537

In another recent research graph theory has been used to discriminate between three
states of healthy controls, patients with MCI due to AD (AD-MCI) and AD patients with
mild dementia attention-deficit disorder (ADD) to evaluate if the abnormality degree could
involve low and/or high degree vertices, the so-called hubs, in both prodromal and over
dementia stages [23]. The mean GC magnitudes across subjects for all links in control,
AD-MCI and ADD groups can be seen in the vertex-to-vertex graph of Figure 15.7.
The measures are in terms of the number of edges (degree), the number of inward (in-
degree), and outgoing edges (out-degree). These parameters were statistically compared
to evaluate the degree of abnormality.

15.3 Motor Neuron Disease

Motor neuron disease usually refers to ALS, but it can also refer to other kinds of neurode-
generative disease that affect the motor neurons, such as progressive primary lateral scle-
rosis, progressive muscular atrophy, and progressive bulbar palsy.
ALS is another heterogeneous neurodegenerative disease characterized primarily by
degeneration of upper and lower motor neurons with variable degrees of extra-motor
involvement. Clinical manifestations of ALS dichotomize into those associated with appar-
ently pure motor system degeneration involving disruption in motor cortex, corticospinal
tracts, and motor networks [24, 25] and degeneration of extra-motor regions, associated
with clinical features of cognitive decline, ranging from mild executive impairment to beha-
vioural variant FTD [26]. While there is an urgent need for noninvasive biomarkers that
address disease heterogeneity, the majority of imaging [24] and electrophysiological [27]
studies to date have focused primarily on quantification of the selective structural degen-
eration and functional deficiencies of motor pathways. The authors in [26] have indicated
that the increasing involvement of broader motor and nonmotor regions and networks
requires a more extensive assessment of large-scale brain connectivity.
In [26] the spectral coherency defined in Eq. (15.1), as a brain connectivity measure, fol-
lowed by a linear discriminant analysis (LDA) classifier has been applied to the EEG data
recorded from approximately 100 ALS patients over eight frequency bands.
Based on their study they demonstrated the validity of spectral EEG as a measure of struc-
tural degeneration in ALS. They also suggested that the increased connectivity (coherences)
reflects network over-activity in affected motor regions (intercortical theta band) and
less degenerated regions such as parietal and frontal areas (gamma band), which likely
represents an indirect potentially compensatory effect of degeneration [26].

15.4 Parkinson’s Disease

PD is the second most common neurodegenerative disorder after AD [28]. It usually affects
older people over the age of 50 and results in tremor as its most common symptom. PD is
caused by the loss of dopaminergic cells in the substantia nigra which is part of the basal
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
538 15 EEG Analysis of Neurodegenerative Diseases

Control
Fp1
Fp2
F7
F3
Fz
F4
F8
T3
C3
To vertex

Cz
C4
T4
T5
P3
Pz
P4
T6
O1
O2
Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6 O1 O2

From vertex

AD-MCI
Fp1
Fp2
F7
F3
Fz
F4
F8
T3
C3
Cz
C4
T4
T5
P3
Pz
P4
T6
O1
O2
Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6 O1 O2

ADD
Fp1
0.02
Fp2
F7
0.018
F3
Fz 0.016
F4
F8 0.014
T3
C3 0.012
Cz
C4 0.01
T4
T5 0.008
P3
Pz 0.006
P4
T6 0.004
O1
0.002
O2
Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6 O1 O2

Figure 15.7 Mean GC magnitudes across subjects for all links in control, AD–MCI and ADD groups. In
the matrix representation ‘from vertex to vertex’ indicates the direction of the information transfer
between electrode pairs. Electrodes are shown from the left to right side, anteriorly–posteriorly. The
colour bar indicates the magnitude of the GC connections [23].
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.4 Parkinson’s Disease 539

ganglia that is partially responsible for motor actions. Symptoms of the disease are not evi-
dent until most of the dopaminergic cells are inactive.
It is characterized by motor symptoms such as tremor, rigidity, slowness of movement,
and problems with gait. Motor symptoms are often accompanied with fatigue, depression,
pain, and cognitive problems.
Deep brain stimulation has become a common practise in stimulating the dopamine
transmitter neurons within the brain to suppress the (mainly resting) tremor.
Monajemi et al. [29] developed a network-based approach for studying the relation
between the tremor intensity and the brain connectivity of patients with PD was introduced.
They introduced an adaptive multitask diffusion strategy to estimate the underlying model
between the gait information and the EEG signals.
In their design they incorporated an S-transform-based connectivity measure [30] in the
multitask diffusion strategy to model the relation between tremor and the brain signals as:

ST
max Im C lk t, f ,0
SC
alk t = 15 11
ST
max Im Cnk t, f ,0
n Nk

ST
where Clk t, f represents the combination weights obtained using the imaginary part
of S-coherency (ImSCoh), Nk is the neighbourhood of node k and Im( ) represents the
imaginary part of a complex-valued entity.
SC
Although using alk as the combination weights incorporates coherency information, it
is an over-simplification of the problem. However, since the brain network is a multitask
complex network where different regions are associated with different tasks, the authors
have proposed a multitask diffusion strategy which incorporates both task-related
objectives as well as coherency information. Therefore, to achieve a robust estimation of
the combination weights in their diffusion adaptation model and reduce the effect of
outliers, they combined the above connectivity values to a set of task-related weights in
multitask scenarios defined as:

w k t − 1 − ψl t 2
MT l Nk
alk t ≈ n N k w k t − 1 − ψn t
2
15 12
0 Otherwise

where wk(t) is the adaptive cooperative filter parameter vector and ψl(t) is the intermediate
vector of parameters, in the following way [5]:
MT SC
λalk t + 1 − λ alk t
l Nk
alk t ≈ n Nk
MT
λank t + 1−λ
SC
ank t 15 13
0 Otherwise

where 0 ≤ λ ≤ 1 and has to be set empirically. This enhanced brain connectivity measure
represents the time–space brain variation relation to the hand tremor.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
540 15 EEG Analysis of Neurodegenerative Diseases

0.12

0.1

0.08
WDCave
k

0.06

0.04

0.02

0
Fp1
Fp2
AFz
F7
F3
Fz
F4
F8
FC5
FC1
FC2
FC6
T7
C3
Cz
C4
T8
CP5
CP1
CP2
CP6
P7
P3
Pz
P4
P8
O1
O2
Node index
Mean WDC in high tremor group Mean WDC in low tremor group

Figure 15.8 Results of the multitask diffusion adaptation method in [5] for high and low Parkinson’s
disease tremor level groups in terms of averaged WDC for λ = 0.7.

They also obtained the so-called weighted degree centrality (WDC) which is a measure
that has been widely used in brain network studies [31–34]. This measure defined as:
N
WDCk t = akl t 15 14
l=1

evaluates the centrality or importance of a node, k, at time instant t in the network according
to its connectivity strength to all the other nodes of the network.
The results show how the differences between the connectivity values of patients with
mild and severe hand tremor are most distinguishable from their EEGs when using the
proposed method. Figure 15.8 shows the WDC values for different EEG channels (nodes)
in low and high tremor patients for an empirically good value of λ = 0.7.
In another research study, the MEG electrodes were clustered into 10 major cortical
regions (as depicted in Figure 15.9) and the data analysis involved calculation of three
synchronization likelihood (SL, a general measure of linear and nonlinear temporal corre-
lations between time series) measures which reflect functional connectivity within (local)
and between the cortical regions in five frequency bands [35].
In this research it has been concluded an increased resting-state functional connectivity
in PD patients compared with the control group. In early-stage PD, the increase is around
the lower alpha range, but with disease progression, the functional connectivity progres-
sively increases over a wider range of 4–30 Hz. They also confirm the association between
beta band coupling and severity of parkinsonism, in particular bradykinesia, and found
some evidence for a similar association between functional connectivity in the theta band
and motor symptoms, in particular tremor. As another finding in this study, cognitive
perseveration in early-stage PD is positively associated with increased inter-hemispheric
functional connectivity in the lower alpha (8–10 Hz) range [35].
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.5 Huntington’s Disease 541

Figure 15.9 Ten brain regions used to

estimate the functional connectivity using
MEG [35].

LF RF

LT LC RC RT

LP RP

LO RO

15.5 Huntington’s Disease

Like PD, Huntington’s disease (HD) is also a neurodegenerative disease and affects the basal
ganglia. This disease can be passed on from parent to child and leads to severe movement
problems.
HD is a result of abnormal protein build-up in the brain, which results in neuronal death.
Individuals with this disease show excessive and undesired movements. Gradually, as the
disease progresses, people with HD will find it increasingly hard to control their movements
and achieve their goals.
Based on a review by Leuchter et al. [36], a major focus in development of novel therapies
for HD is identification of treatments that reduce the burden of mutant huntingtin (mHTT)
protein in the brain. In order to identify and test the efficacy of such therapies, it is essential
to have biomarkers that are sensitive to the effects of mHTT on brain function to determine
whether the intervention has been effective at preventing toxicity in target brain systems
before onset of the clinical symptoms. Quantitative EEG measures brain oscillatory activity
that is regulated by the brain structures that are affected by mHTT in premanifest and early
symptom individuals.
Based on this review a number of researchers have examined the changes in human EEG
as the result of HD. Lazar et al. [37] used polysomnography (PSG) including EEG and con-
cluded that around the 5–7 Hz range (theta and alpha–theta border) the relative power in
pre-HD and early HD is lower during rapid eye movement (REM) compared to controls.
Also, the relative power within the 6–7 Hz range (theta and alpha–theta border) is lower
in pre-HD and early HD during non-rapid eye movement (NREM) relative to controls.
Piano et al. used EEG source localization using low-resolution brain electromagnetic
tomography (LORETA) and discovered that the alpha power is higher in HD during NREM
relative to control and alpha power is lower in HD during REM relative to control [38]. They
also found that the beta power is lower in HD during NREM relative to control and the theta
power is lower in HD during NREM and REM relative to control. Finally, they realized that
the delta power is higher in HD during wake compared to control.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
542 15 EEG Analysis of Neurodegenerative Diseases

Ponomareva et al. [31] utilized quantitative EEG and concluded that within the 8–9 Hz
(alpha) range, the relative power in pre-HD is lower relative to control. They also found that
in the 7–8 Hz (alpha–theta border) range the relative power in pre-HD is lower relative to
control.
EEG-based LORTETA was also used for HD detection by Painold et al. who discovered
that compared to a healthy subject, for an HD patient, the alpha, beta, and theta LORETA
powers are lower, whereas the delta power is higher [32]. Another quantitative EEG
measure also revealed the same results for delta and alpha bands [33].

15.6 Prion Disease

Prion diseases are a group of progressive neurodegenerative diseases that are caused by
misfolded proteins, referred to as prions. During post-translational modification of proteins,
prions act as a folding template, converting proteins into infectious prion form. The
best-known human prion disease is Creutzfeldt–Jakob (mad cow) disease (CJD).
The EEG patterns for CJD are very distinctive and different from the normal EEG
patterns during various states of the brain. Figure 15.10 shows a sample of EEG data for
a CJD patient.
CJD is a long-latency infection caused by a prion. There are four types of CJD, namely
sporadic, variant, familial or inherited, and iatrogenic CJD. Sporadic CJD (also called as
classical CJD) is the most popular type of CJD which is shared between human and animals

FP1 - A1

FP2 - A2

F3 - A1

F4 - A2

C3 - A1

C4 - A2

P3 - A1

P4 - A2

O1 - A1

O2 - A2

1 second

Figure 15.10 A sample of 10-channel EEG of a CJD patient showing clear periodic spikes and diffuse
delta (more on the right frontal, i.e. over FP2 and F4). The discharges share some characteristics of
triphasic waves and therefore could represent rapidly progressive AD.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.6 Prion Disease 543

FP1 - A1
F3 - A1
C3 - A1
P3 - A1
O1 - A1
F7 - A1
T3 - A1
T5 - A1
Fz - A1
Cz - A1
Pz - A1
FP2 - A2
F4 - A2
C4 - A2
P4 - A2
O2 - A2
F8 - A2
T4 - A2
T6 - A2

Figure 15.11 EEG in the very early stages of CJD, showing right-lateralised periodic triphasic
sharp waves.

and the cause is unknown. Variant CJD which is likely to be caused by consuming meat
from a cow that had bovine spongiform encephalopathy (BSE), or ‘mad cow’ disease, is
a similar prion disease to CJD. Familial CJD is a very rare genetic condition where one
of the genes a person inherits from their parent (the prion protein gene) carries a mutation
that causes prions to form in their brain during adulthood, triggering the symptoms of CJD.
Finally, iatrogenic CJD arises from contamination with tissue from an infected person, usu-
ally as the result of a medical treatment or surgical operation.
The characteristic EEG shows biphasic or triphasic discharges that are initially sporadic
and may even be asymmetric (see Figure 15.11). As the disease advances, the pattern
becomes generalized and synchronous with continuous periodic stereotypic 200- to 400-
millisecond sharp waves occurring at intervals of 0.5–1.0 seconds.
The investigation in [34] reveals that in sporadic CJD, the EEG exhibits characteristic
changes depending on the stage of the disease, ranging from nonspecific findings such
as diffuse slowing and frontal rhythmic delta activity (FIRDA) in early stages to disease-
typical periodic sharp wave complexes (PSWC) in middle and late stages of coma (around
30–16 days before death) in preterminal EEG recordings. PSWC, either lateralized (in ear-
lier stages) or generalized, occurs in about two thirds of patients with sporadic CJD, with a
positive predictive value of 95%. PSWC tend to disappear during sleep and may be attenu-
ated by sedative medication and external stimulation. Conversely, seizures may occur in
less than 15% of patients with sporadic CJD. In patients with iatrogenic CJD, PSWC is usu-
ally present with more regional EEG.
Similarly, in some recent studies [39] EEG showed slowing and periodic lateralized dis-
charges over the right hemisphere with triphasic morphology, less often involving the left,
reflecting clinical asymmetry.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
544 15 EEG Analysis of Neurodegenerative Diseases

The above waveform patterns are easy to recognize and therefore, classification of CJD
from normal and many degenerative brain diseases using EEG signals is often achieved
with very high accuracy. This makes the use of EEG for monitoring the disease progress
as a diagnostic tool or as the complementary information to behavioural, pathological,
and histological findings for CJD favourable.
As an example, in [40] the authors used EEG to differentiate patients with early-stage
CJD from rapidly progressive dementia. They used time–frequency representation of the
EEG signals through the continuous wavelet transform (CWT) and permutation entropy
(PE), as complexity measure. They also used an autoencoder neural network for
dimensionality reduction. They also examined support vector machines (SVM) and fully
connected multilayer perceptron (MLP-neural networks [NN]) for their supervised
classification.

15.7 Behaviour Variant Frontotemporal Dementia

Behaviour variant frontotemporal dementia (bvFTD) is characterized by prominent

changes in personality and behaviour that often occurs in people in their 50s and 60s
but can develop earlier or later. In behaviour variant FTD, the nerve cell loss is most
prominent in areas that control conduct, judgement, empathy, and foresight, among other
abilities.
In [41] the authors evaluated the resting-state connectivity in patients with beha-
vioural variant FTD using an SVM. For estimating the brain connectivity, they used a
connectivity measure called weighted symbolic mutual information (wSMI) defined in
[42] which is basically dependent on the similarity of the wave segments between the
nearby channels. This evaluates the extent to which two EEG signals present non-
random joint fluctuations, suggesting that they share information and have three main
properties. First, it looks for qualitative or ‘symbolic’ patterns of increase or decrease in
the signal, which allows a fast and robust estimation of the signals’ entropies. The sym-
bolic transformation depends on the length of the symbols (k = 3 in their application) and
their temporal separation (t = 4, 8, 16, or 32 ms in their application) which follows the PE
defined in [43]. Second, wSMI makes few hypotheses on the type of interactions and
detects the nonlinear couplings. Third, the wSMI weights discard the spurious correla-
tions between EEG signals arising from common sources and favour nontrivial pairs of
symbols.
They found that there was reduced information sharing mainly across mid- and long-
range frontotemporal hubs in bvFTD compared to controls. Using these results an
enhanced classification of bvFTD patients from controls based on neuropsychological data
was achieved. These features offered the best discrimination between bvFTD and AD
patients too. These findings suggest that EEG-derived connectivity analyses could offer
important contributions to the ongoing quest of massively available, inexpensive biomar-
kers for bvFTD. Relative to controls, bvFTD patients exhibited hypoconnectivity between
frontal and temporal regions of interest (ROIs).
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15.9 Summary 545

15.8 Lewy Body Dementia

LBD, also called dementia with Lewy bodies (DLB), is the second most common type of
progressive dementia after AD. Protein deposits, called Lewy bodies, develop in nerve cells
in the brain regions involved in thinking, memory, and movement (motor control). DLB
causes a progressive decline in mental abilities. People with DLB may experience visual
hallucinations and changes in alertness and attention. Other effects include PD-like signs
and symptoms such as rigid muscles, slow movement, and tremors.
Investigations have shown that a combination of quantitative EEG measures (mainly the
EEG power in different conventional frequency bands) and the functional connectivity can
be used to distinguish between DLB, control, and AD subjects [44].
Also, in [45] the authors combined the spectral domain information with fMRI functional
connectivity to compare the resting state in DLB, AD, and control.
The authors in [46] concluded well established slowing of the EEG in the LBD groups
compared to healthy controls and AD patients. Although they did not find higher dominant
frequency variability (DFV) in DLB patients compared to controls as expected, theta DFV
and slow-theta frequency prevalence were positively correlated with cognitive fluctuations
(CF) as measured by clinician assessment of fluctuations. This DLB specific correlation
suggests that a slower and more temporally variable dominant frequency specifically relates
to the CFs seen in DLB and could reveal differential mechanisms underlying CFs in demen-
tia subtypes. Another finding was a significantly higher DFV in AD patients compared to
the other groups. Exploratory analysis showed that quantitative EEG measures could
predict a DLB versus an AD diagnosis with high accuracy, sensitivity, and specificity.

15.9 Summary

People all over the world are suffering from neurodegenerative diseases, which are illnesses
that lead to cell death in the brain. Some neurodegenerative diseases, like PD and HD, affect
the basal ganglia and lead to movement difficulties. Other diseases cause more widespread
cell death and lead to memory loss, which is seen in AD and LBD. There are also rarer types
of neurodegenerative diseases that were not covered in this chapter.
The basis for the assessment of neurodegenerative diseases is to compare the brain net-
work activity of the patient with that of healthy control. This answers a central question in
cognitive neuroscience stating how cognitive functions depend upon coordinated and
integrated activity of specialized, widely distributed brain regions. There is strong support
that a network perspective on the brain is required in order to understand higher brain
functioning. Therefore, the brain connectivity estimates, explained before in Chapter 8
and used in this chapter, are very effective tools in the analysis of brain state.
Although normal brain rhythms are less affected by these diseases, the brain response to
various stimuli as well as brain connectivity and the transfer of information among various
brain zones are seriously affected. Therefore, these parameters are estimated from the EEGs
to assess the disease progress.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
546 15 EEG Analysis of Neurodegenerative Diseases

The use of stem cells to replace the neurons that have died is expected to be a break-
through in the treatment of neurodegenerative diseases. With this currently in research,
hopefully there will soon be help for people with these diseases. Conversely, the collective
information from various screening modalities can even help predicting the onset of AD. As
an example, recent proposals such as that by Winer et al. [47] suggest that sleep may be a
factor associated with accumulation of two core pathological features of AD, i.e. amyloid-β
and τ proteins. Then, the combined positron emission tomography (PET) measures of amy-
loid-β and τ proteins, EEG sleep recordings (for measuring, e.g. time to spindle peak or the
NREM slow wave (SW) association), and retrospective sleep evaluations to investigate the
potential utility of sleep measures in predicting in vivo AD pathology in male and female
older adults.

References
1 Cornblath, E.J., Robinson, J.L., Irwin, D.J. et al. (2020). Defining and predicting
transdiagnostic categories of neurodegenerative disease. Nature Biomedical Engineering 4:
787–800.
2 Villoslada, P., Baeza-Yates, R., and Masdeu, J.C. (2020). Reclassifying neurodegenerative
diseases. Nature Biomedical Engineering 4: 759–760.
3 Mathuranath, P.S., Nestor, P.J., Berrios, G.E. et al. (Dec. 2000). A brief cognitive test battery to
differentiate Alzheimer’s disease and frontotemporal dementia. Neurology 55 (11):
1613–1620.
4 Daianu, M. et al. (2013). Breakdown of brain connectivity between Normal aging and
Alzheimer’s disease: a structural k-core network analysis. Brain Connectivity 3 (4): 407–422.
5 Granger, C.W.J. (1969). Investigating causal relations by econometric models and cross-
spectral methods. Econometrica 37 (3): 424–438.
6 Dauwels, J., Vialatte, F., Musha, T., and Cichocki, A. (2010). A comparative study of
synchrony measures for the early diagnosis of Alzheimer’s disease based on EEG.
NeuroImage 49: 668–693.
7 Pereda, E., Quiroga, R., and Bhattacharya, J. (2005). Nonlinear multivariate analysis of
neurophysiological signals. Progress in Neurobiology 77: 1–37.
8 Jeong, J. (2004). EEG dynamics in patients with Alzheimer’s disease Clin. Neurophysiology
115: 1490–1505.
9 Rossini, P., Rossi, S., Babiloni, C., and Polich, J. (2007). Clinical neurophysiology of aging
brain: from normal aging to neurodegeneration. Progress in Neurobiology 83: 375–400.
10 Alonso, J., Poza, J., Mañanas, M. et al. (2011). MEG connectivity analysis in patients with
Alzheimer’s disease using cross mutual information and spectral coherence. Annals of
Biomedical Engineering 39: 524–536.
11 Stam, C. et al. (2009). Graph theoretical analysis of magnetoencephalographic functional
connectivity in Alzheimer’s disease. Brain 132: 213–224.
12 Stam, C. (2010). Use of magnetoencephalography (MEG) to study functional brain networks
in neurodegenerative disorders. Journal of the Neurological Sciences 289: 128–134.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
References 547

13 Gómez, C., Stam, C., Hornero, R. et al. (2009b). Disturbed beta band functional connectivity
in patients with mild cognitive impairment: an MEG study. IEEE Transactions on Biomedical
Engineering 56: 1683–1690.
14 Nolte, G., Bai, O., Wheaton, L. et al. (2004). Identifying true brain interaction from EEG data
using the imaginary part of coherency. Clinical Neurophysiology 115: 2292–2307.
15 Sanchez Bornot, J.M., Wong-Lin, K., Ahmad, A.L., and Prasad, G. (2018). Robust EEG/MEG
based functional connectivity with the envelope of the imaginary coherence: sensor space
analysis. Brain Topography 31: 895–916.
16 Stam, C.J., Nolte, G., and Daffertshofer, A. (2007). Phase lag index: assessment of functional
connectivity from multi channel EEG and MEG with diminished bias from common sources.
Human Brain Mapping 28: 1178–1193.
17 Vinck, M., Oostenveld, R., van Wingerden, M. et al. (2011). An improved index of phase-
synchronization for electrophysiological data in the presence of volume-conduction, noise
and sample-size bias. NeuroImage 55: 1548–1565. https://doi.org/10.1016/j.
neuroimage.2011.01.055.
18 Dauwels, J., Vialatte, F., Musha, T., and Cichocki, A. (2010). A comparative study of
synchrony measures for the early diagnosis of Alzheimer’s disease based on EEG.
NeuroImage 49: 668–693.
19 Hoechstetter, K., Bornfleth, H., Weckesser, D. et al. (2004). BESA source coherence: a new
method to study cortical oscillatory coupling. Brain Topography 16: 233–238.
20 Supp, G., Schlögl, A., Trujillo-Barreto, N. et al. (2007). Directed cortical information flow
during human object recognition: analyzing induced EEG gamma-band responses in brain’s
source space. PLoS One 2: e684.
21 Escudero, J., Sanei, S., Jarchi, D. et al. (Aug. 2011). Regional coherence evaluation in mild
cognitive impairment and Alzheimer’s disease based on adaptively extracted
magnetoencephalogram rhythms. Journal of Physiological Measurements 32 (8): 1163–1180.
22 Enshaeifar, S., Sanei, S., Cheong-took, C. (2014). Singular spectrum analysis of P300 for
classification. Proceedings of IEEE International Joint Conference on Neural Networks. IJCNN.
23 Franciotti, R., Falasca, N.W., Arnaldi, D. et al. (2019). Cortical network topology in
prodromal and mild dementia due to Alzheimer’s disease: graph theory applied to resting
state EEG. Brain Topography 32: 127–141.
24 Bede, P. and Hardiman, O. (2014). Lessons of ALS imaging: pitfalls and future directions—a
critical review. NeuroImage Clinical 4: 436–443.
25 Schuster, C., Elamin, M., Hardiman, O., and Bede, P. (2016). The segmental diffusivity profile
of amyotrophic lateral sclerosis associated white matter degeneration. European Journal of
Neurology 23: 1361–1371.
26 Nasseroleslami, B., Dukic, S., and Broderick, M. (January 2019). Characteristic increases in
EEG connectivity correlate with changes of structural MRI in amyotrophic lateral sclerosis.
Cerebral Cortex 29: 27–41. (Oxford Pub.).
27 de Carvalho, M., Chio, A., Dengler, R. et al. (2005). Neurophysiological measures in
amyotrophic lateral sclerosis: markers of progression in clinical trials. Amyotrophic Lateral
Sclerosis 6: 17–28.
28 Dorsey, E.R. et al. (2007). Projected number of people with Parkinson disease in the most
populous nations, 2005 through 2030. Neurology 68 (5): 384–386.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
548 15 EEG Analysis of Neurodegenerative Diseases

29 Monajemi, S., Eftaxias, K., Ong, S.-H., and Sanei, S. (2016). An informed multitask diffusion
adaptation approach to study tremor in Parkinson’s disease. IEEE Journal of Selected Topics in
Signal Processing; special issue on advanced signal processing in brain networks 10 (7):
1306–1314.
30 Stockwell, R.G., Mansinha, L., and Lowe, R.P. (Apr. 1996). Localization of the complex
spectrum: the S-transform. IEEE Transactions on Signal Processing 44 (4): 998–1001.
31 Ponomareva, N., Klyushnikov, S., Abramycheva, N. et al. (2014). Alpha-theta border EEG
abnormalities in preclinical Huntington’s disease. Journal of the Neurological Sciences 344:
114–120. https://doi.org/10.1016/j.jns.2014.06.035.
32 Painold, A., Anderer, P., Holl, A.K. et al. (2011). EEG low-resolution brain electromagnetic
tomography (LORETA) in Huntington’s disease. Journal of Neurology 258: 840–854.
https://doi.org/10.1007/s00415-010-5852-5.
33 Hunter, A., Bordelon, Y., Cook, I., and Leuchter, A.F. (2010). QEEG measures in
Huntington’s disease: a pilot study. PLoS Currents 2: RRN1192. https://doi.org/10.1371/
currents.RRN1192.
34 Wieser, H.G., Schindler, K., and Zumsteg, D. (2006). EEG in Creutzfeldt–Jakob disease.
Clinical Neurophysiology 117: 935–951.
35 Stoffers, D., Bosboom, J.L.W., Deijen, J.B. et al. (2008). Increased cortico-cortical functional
connectivity in early-stage Parkinson’s disease: an MEG study. NeuroImage 41: 212–222.
36 Leuchter, M.K., Donzis, E.J., Cepeda, C. et al. (30 March 2017). Quantitative
electroencephalographic biomarkers in preclinical and human studies of Huntington’s
disease: are they fit-for-purpose for treatment development? Frontiers in Neurology
https://doi.org/10.3389/fneur.2017.00091.
37 Lazar, A.S., Panin, F., Goodman, A.O.G. et al. (2015). Sleep deficits but no metabolic deficits
in premanifest Huntington’s disease. Annals of Neurology 78: 630–648. https://doi.org/
10.1002/ana.24495.
38 Piano, C., Mazzucchi, E., Bentivoglio, A.R. et al. (2016). Wake and sleep EEG in patients with
Huntington disease: an eLORETA study and review of the literature. Clinical EEG and
Neuroscience 48 (1): 60–71. https://doi.org/10.1177/1550059416632413.
39 Ganesh, A., Hoyte, L.C., Agha-Khani, Y., and Yeung, M.M.C. (April 17, 2018). Teaching
NeuroImages: DWI and EEG findings in Creutzfeldt-Jakob disease. Neurology 90 (16):
e1450–e1451.
40 Morabito, F.C., Campolo, M., Mammone, N. et al. (2017). Deep learning representation from
electroencephalography of early-stage Creutzfeldt-Jakob disease and features for
differentiation from rapidly progressive dementia. International Journal of Neural Systems 27
(02): 1650039. 12 pages.
41 Dottori, M., Sedeño, L., Caro, M.M. et al. Towards affordable biomarkers of frontotemporal
dementia: a classification study via network’s information sharing. Nature Scientific Reports
7: 3822. https://doi.org/10.1038/s41598-017-04204-8.
42 King, J.R., Sitt, J.D., Faugeras, F. et al. (2013). Information sharing in the brain indexes
consciousness in noncommunicative patients. Current Biology: CB 23: 1914–1919.
https://doi.org/10.1016/j.cub.2013.07.075.
43 Bandt, C. and Pompe, B. (2002). Permutation entropy: a natural complexity measure for time
series. Physical Review Letters 88: 174102.
 10.1002/9781119386957.ch15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/9781119386957.ch15 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [19/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
References 549

44 van der Zande, J.J., Gouw, A.A., van Steenoven, I. et al. (2018). EEG characteristics of
dementia with lewy bodies, Alzheimer’s disease and mixed pathology. Frontiers in Aging
Neuroscience 10: 190. 10 pages.
45 Schumacher, J., Peraza, L.R., Firbank, M. et al. (2019). Dysfunctional brain dynamics and
their origin in Lewy body dementia. Brain 142: 1767–1782.
46 Stylianou, M., Murphy, N., Peraza, L.R. et al. (2018). Quantitative electroencephalography as
a marker of cognitive fluctuations in dementia with Lewy bodies and an aid to differential
diagnosis. Clinical Neurophysiology 129: 1209–1220.
47 Winer, J.R., Mander, B.A., Helfrich, R.F. et al. (August 7, 2019). Sleep as a potential biomarker
of tau and -amyloid burden in the human brain. The Journal of Neuroscience 39 (32):
6315–6324.