Advances in Sleep EEG Signaling in Alzheimer's Disease Prediction

Yulin Jiang

College of Chemical Engineering, Huaqiao University, Xiamen, 362021, China

Keywords: Alzheimer’s Disease, Sleep EEG, NREM Sleep.

Abstract: With the acceleration of global aging, Alzheimer's Disease (AD) has emerged as a grave public health issue.

At present, the commonly employed diagnostic methods have certain drawbacks. In contrast, sleep

electroencephalography (EEG) signals have garnered significant attention in the area of AD prediction,

mainly because of their non - invasive nature, repeatability, and low cost. In this paper, we review the research

progress of sleep EEG signals in AD prediction, elaborate the pathological mechanisms of AD, compare the

advantages and disadvantages of traditional detection methods, and analyze the current status and

development of sleep stage classification system technology is ongoing. When concentrating on the

connection between non - rapid eye movement (NREM) sleep stages and AD, it has been discovered that in

AD patients, the σactivity shows a decline and the EEG undergoes a slowdown during NREM sleep, and that

σ power during NREM sleep is positively correlated with cognitive ability, which may be used as a reference

standard for AD detection. Future research efforts should be dedicated to optimizing the algorithm in order to

enhance the precision of sleep stage classification, integrate multimodal data to explore the relationship

between sleep and AD, and carry out a large-scale longitudinal study to validate the sleep EEG indexes, so as

to promote the development of early warning and precise intervention for AD.

1 INTRODUCTION

In today's society, with the acceleration of global

aging, Alzheimer's Disease (AD), as a

neurodegenerative disease, is the most common form

of dementia, the third most expensive disease and the

sixth leading cause of death worldwide. It has become

a serious public health challenge.

Following the deposition of insoluble amyloid-β

(Aβ), tau accumulates in neocortical cells, leading to

neuronal cell death, synapse loss, brain volume

reduction, and cognitive impairment. In the absence

of cognitive symptoms, the progression of

Alzheimer's disease (AD) involves the gradual

accumulation of pathological changes, creating a

critical window for timely therapeutic intervention.

Sleep patterns are now emerging as a potential

biomarker for AD pathology and a predictor of future

cognitive decline Lucey, et al.,2019).

AD is difficult to diagnose, and symptoms can be

easily misinterpreted as a normal consequence of

aging, requiring multiple investigations and the

exclusion of other causes. Two significant

pathological changes occur: the deposition of β -

amyloid plaques and the formation of

hyperphosphorylated tau neurofibrillary tangles.

Biomarkers like cerebrospinal fluid (CSF) analysis

and positron emission tomography (PET) imaging,

when integrated with clinical evaluations, are

commonly employed to diagnose the disease, but the

former is an invasive procedure that causes

physiological discomfort to the patient such as the

risk of infection and pain at the puncture site, and the

latter is expensive to perform with expensive

equipment and a high cost of learning, which greatly

limits its popularity. The latter is expensive and costly

to learn, which greatly limits its popularity. On the

other hand, sleep electroencephalography (EEG)

signal acquisition EEG devices are affordable. In

recent years, the progress of sleep stage recognition

technology has been remarkable. Due to the

limitation of manual scoring, the development of

automatic sleep stage classification system (ASSC)

has been accelerated, using the PhysioNet Sleep EDF

database and a decision tree classifier, the model

achieved an average sensitivity of 89.06%, specificity

of 98.61%, and accuracy of 93.13%., which improved

the feasibility and speed of practical application of

ASSC (Lucey, et al.,2019). feasibility and speed of

practical application (Aboalayon, et al., 2016).

Jiang, Y.

Advances in Sleep EEG Signaling in Alzheimer’s Disease Prediction.

DOI: 10.5220/0014386400004933

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Biomedical Engineering and Food Science (BEFS 2025), pages 39-45

ISBN: 978-989-758-789-4

During sleep, the electrical activity of the brain

exhibits a rich variety of regular changes, which

contain a vast amount of information regarding the

state of brain health. Sleep is mainly classified into

two stages: Rapid Eye Movement (REM) sleep and

Non - Rapid Eye Movement (NREM) sleep, and each

of them has unique electrical characteristics and

physiological functions, and may play different but

interrelated roles in the development of AD.

Given the unique advantage of sleep EEG signal in

reflecting the functional state of the brain, as well as

its noninvasive, reproducible, and relatively low-cost

features, investigating the use of sleep EEG signals

for AD prediction not only broadens our

comprehension of AD pathogenesis but also offers

innovative approaches for its early detection. The

present study aims to review the research progress of

sleep EEG signaling in AD prediction in recent years,

focusing on the changes of EEG characteristics

during REM and NREM sleep stages and their

correlation with the pathophysiological process of

AD, evaluating the strengths and weaknesses of

current research methods and exploring future

directions, this study aims to establish a solid

theoretical foundation and practical guidance for

early AD detection and targeted intervention.

2 AD PATHOGENESIS

2.1 Mechanisms of Aβ and Tau in the

Induction of AD

Aβ is produced by cleavage of amyloid precursor

protein (APP) by β - secretase and γ - secretase.

Under normal conditions, Aβ can be cleared, but in

AD patients, there is an imbalance between the

production and clearance of Aβ, leading to the

abnormal deposition of Aβ in the brain and the

formation of senile plaques. Aβ oligomers have

neurotoxicity, which can bind with receptors on the

cell membrane of neurons, these disruptions impair

neural signaling and synaptic plasticity, while

simultaneously promoting the generation of reactive

oxygen species. This leads to oxidative stress, which

damages cell membranes, proteins, and mitochondria,

ultimately causing the death of neurons.

Normally, tau protein promotes the assembly of

microtubules and maintains their stability to ensure

intra-neuronal substance transportation. In AD, tau

protein is abnormally hyperphosphorylated and its

ability to bind to microtubules decreases, resulting in

microtubule depolymerization, which destroys the

cytoskeleton structure of neurons, affects axonal

transport, and prevents neurons from taking up

nutrients and transmitting signals normally. Over-

phosphorylated tau protein aggregates to form

neurogenic fiber tangles. These tangles accumulate in

neurons, hindering normal physiological activities of

neurons, and can spread among neurons, accelerating

neurodegeneration.

2.2 Other Relevant Pathological

Factors and Interactions

Aβ deposition, tau protein abnormalities, impaired

mitochondrial function, and cerebrovascular

pathology all play key roles in the complex

pathogenesis of AD.

Aβ is produced by cleavage of APPs by specific

enzymes and is normally cleared. In AD patients, the

balance between Aβ production and clearance is

disrupted, leading to its excessive accumulation in the

brain. tau proteins become abnormally

hyperphosphorylated, disrupting their ability to bind

to microtubules and interfering with intra-neuronal

transport of substances. the deposition of Aβ and the

abnormalities of the tau proteins activate microglial

cells and astrocytes. Activated microglia release pro -

inflammatory cytokines, such as interleukin - 1β (It

seems there might be a mistake in your original

"interleukin - 1 Aβ", perhaps you meant interleukin -

1β) and tumor necrosis factor - α, while astrocytes

expand in reaction to inflammatory signals, releasing

a variety of cytokines and chemokines, which triggers

neuroinflammation and damage to neuronal cells.

Mitochondrial function is also impaired in the brain

of AD patients. Mitochondrial dysfunction can trigger

oxidative stress, partially clarifying the intricate

mechanisms behind oxidative damage in AD. Beyond

ATP production, mitochondria play a key role in

controlling cell death by storing various apoptotic

factors, which are released during apoptosis. In AD

patients, mitochondrial impairment, elevated

oxidative stress, and neuronal apoptosis have been

observed (Moreira, et al., 2012). These findings imply

that mitochondrial malfunction could be the impetus

behind neuronal degeneration and demise in AD.

Disruptions in the mitochondrial respiratory chain

and impaired electron transport lead to a decline in

membrane potential and diminished energy

generation. At the same time, reactive oxygen species

production increases, exceeding the cellular

antioxidant capacity and oxidatively damaging lipids,

BEFS 2025 - International Conference on Biomedical Engineering and Food Science

proteins, and other biomolecules, making neurons

more susceptible to oxidative stress damage.

Certain gene mutations are closely related to AD,

such as mutations in APP, progerin 1 (PS1) and

progerin 2 (PS2), which can lead to familial AD; and

the ε4 allele of the apolipoprotein E (APOE) gene

significantly heightens the risk of developing

sporadic AD (Scheltens, et al., 2021).

Cerebrovascular lesions are equally important in

the development of AD. Chronic hypertension also

impairs the integrity of the blood-brain barrier (BBB),

leading to cerebral edema and the introduction of

systemic elements into the brain parenchyma, and

chronic hypertension also impairs the integrity of the

BBB, resulting in brain swelling and the infiltration

of systemic components into brain tissue (Santos, et

al., 2012). Cerebrovascular endothelial dysfunction

affects vasodilatation and vasoconstriction, resulting

in reduced cerebral blood flow, inadequate nutrient

supply to brain tissue, and accumulation of metabolic

wastes. The damage to the blood-brain barrier is even

more serious, its permeability increases, harmful

substances enter the brain tissue, triggering

inflammation and immune damage, and it also affects

the removal of Aβ, prompting the further deposition

of Aβ and accelerating the development of AD

disease. These pathologic processes interact with

each other and jointly promote the development and

deterioration of AD.

3 PROGRESS IN DETECTION

RESEARCH

3.1 Traditional Testing Methods and

Limitations

3.1.1 Mini-Mental State Examination

(MMSE)

The MMSE is a widely used clinical instrument for

AD detection. This 30-question test evaluates

cognitive abilities, including attention, orientation,

memory, calculation, language, and visuospatial

skills, such as drawing complex shapes (Arevalo, et

al., 2012). The MMSE score offers a quantitative

measure of cognitive decline in older adults, aiding

doctors in diagnosis and treatment planning.

However, the assessment dimensions of MMSE are

limited, mainly focusing on several major aspects of

cognitive function, and the assessment of some

complex cognitive functions, such as executive

function and social cognition, is not comprehensive

enough. If the elderly has a high level of education,

there may be cases where the MMSE score is still in

the normal range even though there is some cognitive

impairment, thus masking the condition. For some

specific cognitive dysfunctions, such as executive

dysfunction, the MMSE may not be able to detect

them accurately, which may lead to an incomplete

assessment of the patient's cognitive function.

Patients with AD may develop these impairments

during the course of the disease, but the difficulty of

detecting them on MMSE may affect the overall

judgment of the patient's condition.

3.1.2 CSF

Biomarkers in CSF can directly reflect the

pathophysiologic process of AD in the brain.

Pathologic changes, such as Aβ deposition, may

occur in the brain before the onset of clinical

symptoms of AD, and biomarker levels in the CSF

may change accordingly. The most intensively

investigated biomarkers of Alzheimer's disease (AD)

are the cerebrospinal fluid proteins that are

pathologically related, namely β - amyloid 42 (Aβ

1 - 42), total tau (t - tau), and tau phosphorylated at

amino acid 181 (p - tau181). Many laboratories use

enzyme-linked immunosorbent assays (ELISA) to

detect these proteins (Wang et al., 2012). By

immobilizing an antibody that specifically recognizes

Aβ42 on a solid-phase carrier and adding it to a CSF

sample, the Aβ42 in the sample will bind to the

antibody, and then an enzyme-labeled secondary

antibody will be added, which will produce a color

change through the reaction between the enzyme and

the substrate, and then the absorbance will be

measured by using an enzyme marker, and compared

with a standard curve. If the level of Aβ42 in the

CSF decreases significantly, it suggests that

Alzheimer's disease may be present. However, the

CSF test requires lumbar puncture to obtain CSF,

which is an invasive operation that may bring some

pain and risk to patients, and the CSF test involves

special testing equipment, reagents, and specialized

technicians, and the overall cost is relatively high,

which may bring some financial burden to patients

and the health insurance system, and to a certain

extent, limit its wide application.

3.1.3 PET

PET technology allows for the evaluation of various

functional processes in the brain of AD patients

Advances in Sleep EEG Signaling in Alzheimer’s Disease Prediction

during their survival. This method allows for the 3D

visualization and quantification of metabolic (glucose

metabolism) and neurotransmitter activity. It also

provides insights into the pathological mechanisms of

AD. PET scans enable clinicians to visually analyze

results through color coding and, crucially, gather

quantitative data on brain regions. This data supports

objective evaluation of diagnostic precision and

treatment outcomes. PET can identify early metabolic

and pathological brain changes before noticeable

clinical symptoms appear. With specific tracers, such

as the glucose analog of brain glucose metabolism, 2-

[18F]-fluoro-2-deoxygenase, PET is able to detect

subtle metabolic and pathological changes in the

brain before they become clinically apparent.

Oxygen-d-glucose (18F-FDG) can be used to

monitor cerebral glucose metabolism (Nordberg,et

al., 2012). This tracer has been widely used in

radiopharmaceutical imaging studies and clinics of

AD, which can clearly show the metabolic or

pathological changes in different regions of the brain

and help doctors accurately determine the site and

extent of lesions. In AD diagnosis, it can clarify the

functional abnormality of brain areas closely related

to cognitive function, such as hippocampus, internal

olfactory cortex, etc., which can provide an important

basis for localized diagnosis of the disease and

evaluation of the disease, and help to differentiate it

from other diseases that may lead to cognitive

disorders. PET test not only shows the anatomical

structure of the brain, but also more importantly

reflects the functional state of the brain, such as the

metabolic activity of the neurons, neurotransmitter

changes and so on, neurotransmitter changes, etc.

However, the PET test itself is expensive, and with

the cost of the tracer, the overall cost of the test is

usually high. PET equipment is expensive, with high

maintenance costs and high requirements for

installation environment and technicians, resulting in

its limited popularity in medical institutions. At the

same time, the analysis and interpretation of PET

images require specialized nuclear medicine doctors

or specially trained personnel who are not only

familiar with the normal anatomy and physiological

functions of the brain, but also understand the

characteristics of PET performance in various disease

states.

Therefore, in Alzheimer's disease detection, EEG

has outstanding advantages over mainstream

methods. Firstly, it is non-invasive. CSF requires

lumbar puncture, which is risky, while EEG only

places electrodes on the scalp. Secondly, it has a

higher detection accuracy and can capture early

abnormalities in neuronal electrical activity.

Furthermore, in terms of economy and popularity,

CSF and PET testing equipment and process costs are

high, while EEG equipment is cheap, with low

learning costs, and can be operated by primary

healthcare professionals after short-term training,

which is more conducive to popularization, and more

patients can benefit from early diagnosis, which has a

great potential for the detection of AD.

3.2 State of the Art and Development

of Sleep Stage Classification System

Technology

Classifying sleep stages is essential for studying

sleep, diagnosing sleep disorders, and assessing

treatments. It enhances our understanding of sleep

mechanisms and offers a foundation for managing

sleep-related conditions. At present, the sleep stage

classification system technology presents diverse

characteristics in methods and applications, and also

faces many challenges, and the future development

direction is becoming clearer.

Sleep specialists usually perform manual sleep

stage scoring through the analysis of

neurophysiological signals gathered in sleep

laboratories. This process is often challenging,

monotonous, and time-intensive. Scoring is usually

based on polysomnographic (PSG) data recorded

during overnight hospital stays. In traditional

practice, overnight PSG recordings consist of EEG,

electrooculogram (EOG), electromyogram (EMG),

and electrocardiogram (ECG) data. These recordings

are manually assessed by sleep specialists based on

the 1968 guidelines established by Rechtschaffen and

Kales (R&K) (Konkoly, et al., 2012). PSG recordings

are divided into 20- or 30-second intervals and

classified into wakefulness (W), REM sleep, and

NREM sleep. Due to their multi-channel signals and

expert-based visual analysis, PSG remains the gold

standard for assessing sleep in laboratory studies.

Polysomnography offers comprehensive insights into

sleep architecture, duration, and quality. However, it

is costly, labor-intensive, and unsuitable for field

applications, as it requires a sleep technician to install

equipment and place multiple electrodes on the face

and scalp (Arevalo, et al., 2012).

Consequently, the process of sleep stage scoring

incurs high costs, is prone to human mistakes, and is

frequently tiresome and demands a significant

amount of time. Analyzing overnight sleep recordings

usually requires 2 to 4 hours, and in some studies,

there has been a 90% expert agreement on sleep stage

BEFS 2025 - International Conference on Biomedical Engineering and Food Science

classification (Konkoly, et al., 2012). In addition,

sleep stage scoring using PSG usually requires a

hospital setting where subjects have to wait on a

waiting list for some time. Due to the limitations of

manual sleep stage scoring, there is an increasing

demand for the development of automated sleep stage

classification systems (ASSC).

The research divides dual-channel EEG signals

into quasi-steady-state segments, extracts features

using Short-Time Fast Fourier Transform (STFT),

reduces dimensionality with the fuzzy C-Means

algorithm, and constructs an ASSC system

employing a multi-class SVM. The system achieved

an accuracy of 70.92% (Al-Aloqaly, et al., 2012). To

improve the classification accuracy, the researchers

tried to fuse multiple physiological signals. In

addition to EEG, EOG, and EMG, signals such as

heart rate, respiratory rate, and oxygen saturation

were incorporated. Multimodal data fusion can

provide more comprehensive sleep information,

adopt more advanced algorithms, study new deep

learning architectures, such as models based on the

attention mechanism, which can better focus on key

sleep signal features; and explore the application of

generative adversarial networks (GANs) in sleep data

augmentation and model optimization, to improve

model performance. We will further improve the

multimodal data fusion method, combine artificial

intelligence and big data analysis technology, mine

the complex relationship between sleep data, and

realize the comprehensive assessment of sleep

quality, early warning of sleep disorders, and the

formulation of personalized treatment plans.

3.3 Association between NREM Sleep

Stages and AD and Research

Progress

3.3.1 The Relationship between non Rapid

Eye Movement (NREM) and AD

Starting from the prodromal phase of AD, patients

exhibit slower EEG rhythms while awake, potentially

linked to poor sleep quality. To explore the

connection between arousal and sleep, we analyzed

EEG activity during sleep, as well as before and after

sleep, in patients suffering from Alzheimer's disease

(AD), those with mild cognitive impairment (MCI),

as well as healthy individuals used as controls. It was

found that individuals with AD, as well as those

suffering from mild cognitive impairment, presented

a longer sleep latency and less slow-wave sleep. The

NREM sleep phase is typically characterized by

reduced σ activity, which reflects the absence of

the sleep spindle. For both AD and MCI patients,

EEG slowing is characteristic of REM sleep and

wakefulness, and there is a strong correlation between

these two phenomena, suggesting a common

neuropathological mechanism.

Furthermore, EEG changes from evening to early

morning during wakefulness revealed a gradual

reduction in nocturnal δ activity in both MCI and

AD patients. This suggests a progressive decline in

the restorative effects of sleep on circadian rhythms,

aligning with the impaired high-frequency sleep

activity observed in AD patients.

In this process, NREM stage sleep is crucial for

memory consolidation. It plays a facilitating role in

transforming short - term memory into long - term

memory. Moreover, it is of great significance in

maintaining learning ability and cognitive functions.

3.3.2 Power of EEG During NREM Sleep

The histograms of the spectral power at cortical sites

and bands from Figure1 show the power of the two

groups in the α and σ bands. Looking at the graph as a

whole, there is a difference in the α and σ band power

between the AD and HC groups at different cortical

sites. The α and σ bands were chosen because the σ

band is associated with the relaxation and attentional

states of the brain, and in AD patients, altered brain

function may affect their relaxation and attentional

regulation mechanisms. σ band is associated with the

sleep spindle wave, which is critical for memory

consolidation, and AD patients with impaired memory

may have characteristic changes in this band. Other

frequency bands, such as the δ band, are potentially

related to AD, but the two more distinguishable

frequency bands are discussed briefly here. α and σ

power in AD patients at T3 and T5 correspond to the

temporal lobe region of the brain, and the reduced α

and σ power at these locations may indicate abnormal

neuronal activity in the temporal lobe region. α power

is reduced, reflecting the impaired function of

relaxation and attentional regulation of the brain, and

the reduced α power suggests a weakening of the

activity of sleep spindles and impairs memory

consolidation. Reduced α power suggests that sleep

spindle wave activity is impaired, affecting memory

consolidation. This difference may reflect the

alteration of cortical function in AD patients, which to

some extent provides data support for the study of the

neurophysiological mechanism of AD, and helps to

further explore the characteristics and patterns of the

abnormalities in the brain function of AD patients.

Advances in Sleep EEG Signaling in Alzheimer’s Disease Prediction

Figure 1: Spectral Power Histograms of Cortical Sites and Frequency Bands in AD (Alzheimer's Disease) and HC (Healthy

Control) Groups.

3.3.3 Correlation between σ Power and

Cognitive Ability

From Table 1, it is possible to evaluate the association

between the level of cognitive ability and the σ power

(sigma power) during NREM sleep. The graph clearly

shows that there is a significant correlation (p ≤

0.0054) between MMSE scores and σ power during

NREM sleep at different EEG loci (O1, O2, P3, Pz,

T5, T6) (D’Atri, et al., 2012). The correlation

coefficients r values ranged from 0.28 - 0.32 and the

p values were extremely small, indicating that this

correlation was highly statistically significant. Based

on this, we can basically conclude that there is a

positive correlation between σ power and cognitive

ability, i.e., the higher the σ power during NREM

sleep, the higher the corresponding MMSE score and

the stronger the cognitive ability.

Table 1: Correlation (Pearson's R) between MMSE Score

and Sigma Power During NREM Sleep, and EEG Slowing

Index During REM Sleep (P ≤ 0.0054) (D’Atri, et al.,

2012).

EEG site σ power in NREM sleep

r p

O1 0.32 0.000077

O2 0.32 0.000055

P3 0.28 0.0006

Pz 0.31 0.000093

T5 0.31 0.00014

T6 0.29 0.00033

Given this association, σ power during NREM

sleep has the potential to be used as a reference

standard for EEG to detect AD levels. In clinical

practice and research, detecting the power in this

frequency band of the EEG may be able to assist in

determining the cognitive state of an individual and

provide valuable information for early screening and

assessment of AD. However, more studies are needed

to further validate its accuracy and reliability.

4 CONCLUSION

With the limitations of traditional AD detection

methods, sleep EEG signaling has become a hot

research topic due to its unique advantages. This

paper comprehensively analyzed the pathological

mechanisms of AD, the advantages and

disadvantages of traditional detection methods, and

the current state of the art of sleep stage classification

system, and focused on the association between

NREM sleep stages and AD, and found that the

cortical functional activities of AD patients differed

from those of healthy controls in terms of α and σ

band power, and clarified the positive correlation

between the σ power and the cognitive ability.

However, the accuracy and reliability of sleep

EEG signal for AD prediction still need to be

improved. In the future, we can focus on conducting

large-scale and multi-center clinical trials to further

validate the relevant indexes, exploring more

frequency bands and brain regions, optimizing the

detection techniques and analysis algorithms to

enhance the precision and steadiness of the

prediction, and promoting the development of AD

early diagnosis and intervention techniques.

REFERENCES

Aboalayon, K. A. I., Faezipour, M., Almuhammadi, W. S.,

& Moslehpour, S. 2016. Sleep stage classification using

EEG signal analysis: A comprehensive survey and new

investigation. Entropy, 18(8), 272.

BEFS 2025 - International Conference on Biomedical Engineering and Food Science

D'Atri, A., et al. 2021. EEG alterations during wake and

sleep in mild cognitive impairment and Alzheimer's

disease. iScience, 24(4), 102330.

Lucey, B. P., McCullough, A., Landsness, E. C.,

Toedebusch, C. D., McLeland, J. S., Zaza, A. M.,

Fagan, AA. M., McCue, L., Xiong, C., Morris, J. C.,

Benzinger, T. L. S., & Holtzman, D. M. 2019. Reduced

nonrapid eye movement sleep is associated with tau

pathology in early Alzheimer's disease. Science

Translational Medicine, 11(474), eaau6550.

Moreira, P. I., Cardoso, S. M., Santos, M. S., & Oliveira, C.

R. 2005. The key role of mitochondria in Alzheimer's

disease. Journal of Alzheimer's Disease, 8(2), 101–110.

Nordberg, A., et al. 2010. The use of PET in Alzheimer

disease. Nature Reviews Neurology, 6(2), 82–94.

Santos, C. Y., et al. 2017. Pathophysiologic relationship

between Alzheimer's disease, cerebrovascular disease,

and cardiovascular risk: A review and synthesis.

Alzheimer's & Dementia: Diagnosis, Assessment &

Disease Monitoring, 7, 69–87.

Scheltens, P., De Strooper, B., Kivipelto, M., Holstege, H.,

Chételat, G., Teunissen, C. E., ... & van der Flier, W.

M. 2021. Alzheimer's disease. The Lancet, 397(10284),

1577–1590.

Wang, L.-S., et al. 2012. Comparison of xMAP and ELISA

assays for detecting CSF biomarkers of Alzheimer's

disease. Journal of Alzheimer's Disease, 31(2), 439–

445.

Advances in Sleep EEG Signaling in Alzheimer’s Disease Prediction