REGIONAL DIFFERENCES IN NEONATAL SLEEP

ELECTROENCEPHALOGRAM

Karel Paul 1), Vladimír Krajča 2), Zdeněk Roth 3), Jan Melichar 1), Svojmil Petránek 2)

1) Institute for the Care of Mother and Child, Prague, Czech Republic

2) Faculty Hospital Bulovka, Department of Neurology, Prague, Czech Republic

3) National Institute of Public Health, Prague, Czech Republic

Karel Paul

Institute for the Care of Mother and Child

14 710 Prague 4

Czech Republic

Tel. : +420 296511498 Fax. : +420 241432572 e-mail: karelpaul@seznam.cz

ABSTRACT

Background and purpose: While EEG features of the maturation level and behavioral states

are visually well distinguishable in fullterm newborns, the topographic differentiation of the

EEG activity is mostly unclear in this age. The aim of the study was to find out wether the

applied method of automatic analysis is capable of descerning topographic particulaities of

the neonatal EEG. A quantitative description of the EEG signal can contribute to objective

assessment of the functional condition of a neonatal brain and to rafinement of diagnostics

of cerebral dysfunctions manifesting itself as “dysrhytmia”, “dysmaturity” or

“disorganization”.

Subjects and methods: We examined polygraphically 21 healthy, full-term newborns during

sleep. From each EEG record, two five-minute samples were subject to off-line analysis and

were described by 13 variables: spectral measures and features describing shape and

variability of the signal. The data from individual infants were averaged and the number of

variables was reduced by factor analysis.

Results: All factors identified by factor analysis were statistically significantly influenced

by the location of derivation. A large number of statistically significant differences was also

found when comparing the data describing the activities from different regions of the same

hemisphere. The data from the posterior-medial regions differed significantly from the other

studied regions: They exhibited higher values of spectral features and notably higher

variability. When comparing data from homotopic regions of the opposite hemispheres, we

only established significant differences between the activities of the anterior-medial regions:

The values of spectral features were higher on the right than on the left side. The activities

from other homotopic regions did not differ significantly.

Conclusion: The applied method of automatic analysis is capable of discerning differences

in the sleep EEG activities from the individual regions of the neonatal brain.

Significance: The capability of the used method to discriminate regional differences of the

neonatal EEG represents a promise for their application in clinical practice.

Keywords: Full-term newborn; EEG; Regional differences; Automatic analysis

INTRODUCTION

When analyzed visually, the EEG activity of sleeping full-term newborns at the first

glance appears topographically non-differentiated for the most part. The EEG atlases

dealing with the earliest age either do not mention the regional differences in the EEG

signal in full-term sleeping newborns at all [1,2], or the EEG activity of these infants is

being described as „uniformly distributed‟ [3]. The reason for this is probably the fact that

the human eye will not discern differences in the activities from the individual cranial

regions. However, the application of computing technology has proven that the

electroencephalogram of a full-term newborn is in fact topographically differentiated. The

regional differences in the values of spectral energies were described [4-8].

Intrahemispheric and interhemispheric coherence of the EEG activity had been studied [8-

14]. Automatic brain mapping was applied to the neonatal EEG [15,16]. Topographic

interdependencies of the neonatal EEG have been examined by the means of non-linear

methods [17-20].

In the present study, we have applied a multi-channel automatic method based on

adaptive segmentation [21] in order to describe the EEG activity from the specific regions

of the neonatal brain. This method evaluates not only spectral measures but also additional

features as amplitude level, shape and variance of the signal, in which it comes close to

visual analysis. The objective of the study is to verify whether the applied method is capable

of discerning the differences in the EEG activities of the specific brain regions. It is possible

to suppose that if the used method is able to discerne physiological regional EEG

differences it will be possible to use the method in a detection and objective description of

topographic deviations in patients with a cerebral pathology.

SUBJECTS

We included 21 healthy full-term newborns in the study. They were born in the 39th

to the 40th

week of gestation, the Apgar score was >7 in the first minute and >8 in the fifth

minute, and their birthweights ranged from 3010 to 3950g. The infants were examined in

the 4th

and 10th

day of their life. Parents of the infants were informed of the methods and

purposes of the examination and gave their consent. The project was approved by the

institute‟s ethical commission.

METHODS

The examinations were carried out in an EEG laboratory in standardized conditions

after morning feeding and lasted 90-120 minutes. The examination room was noise-

protected and background noise level did not exceed 45dB. The illumination level was

reduced to a degree that would enable the observer to just perceive changes in infant‟s

behavior. Room temperature was in the 23-25°C range. Disturbing environmental stimuli

were excluded. Infants were examined in a crib, placed in supine position. The EEG activity

was recorded polygraphically from eight bipolar derivations, positioned under the system

10-20 (Fp1-C3, C3-O1, Fp1-T3, T3-O1, Fp2-C4, C4-O2, Fp2-T4, T4-O2); the reference

derivation, linked ear electrodes; filter setting, 0,2 and 60Hz; sensitivity, 100μV per 10mm.

The respiration (PNG), ECG, EOG, and EMG of chin muscles were also recorded.

Electrode impedances were not higher than 5kOhm. The recording was performed using the

Brain-Quick (Micromed) digital system with sampling frequency of 128Hz and the data

were stored on CDs. An observer continuously recorded any change in infant‟s behavior on

the polygram.

Two five-minute-samples free of artifacts (segments contaminated by artifacts were

eliminated by visual inspection) were selected from the EEG record of each infant. One

sample was chosen from the middle part of quiet sleep, the other from the middle of the

subsequent active sleep. In this study, we have defined mentioned sleep states according to

the following criteria: Quiet sleep was defined as sleep with closed eyes, absence of eye

movements, regular breathing, absence of body movements except for startles, and the

typical EEG pattern „tracé alternant‟. Active sleep was defined as a behavioral state in

which the infant‟s eyes were closed or nearly closed, eye movements were apparent,

breathing was irregular, and mimic muscle movements, small movements of extremities and

even large generalized movements occurred intermittently. The EEG showed the „activité

moyenne„ pattern [3].

Quantitative processing of EEG was performed off-line. Subject to analysis were

data from the above-mentioned bipolar montage. A method based on multi-channel adaptive

segmentation [21] was used. The method was selected for the following reasons: (a) The

algorithm of the adaptive segmentation divides the EEG signal into quasi-stationary

segments of variable length. The idea was that the feature extraction from such relatively

homogeneous epochs would be substantially more effective than the feature extraction from

fixed epochs. This holds especially true when analyzing the highly variable pattern as tracé

alternant. (b) The division of the signal into quasi-stationary segments made it possible to

evaluate length, number and proportional occurrence of these segments and thus to quantify

the stability and variability of the signal.

The method of applied automatic analysis was explained in detail in our previous

paper [22]. Therefore it is described only briefly in this study. Using adaptive segmentation,

the EEG signal from each derivation was divided into relatively homogeneous segments of

variable length. The limits of the segments were in fact defined by the change in stationary

character of the signal. The segments were distributed into three classes according to their

maximum voltage. The segments whose amplitude didn‟t exceed 50μV were placed into the

1st class, the 2

nd class contained segments with voltage higher than 50μV and lower than

90μV, and the 3rd class was occupied by segments with the amplitude of 90μV and more.

Examples of the application of adaptive segmentation and the distribution of segments into

voltage classes are presented in Fig. 1. The activity of each segment was then described by

ten features. The AV feature described the variance of the segment‟s amplitude; Mm defined

the value of the maximum amplitude „peak-to-peak‟; the following five features provided

information about the value of spectral amplitude in five frequency bands, δ1 in the 0.2-

1.5Hz band, δ2 in the 1.6-3Hz band, θ1 in the 3.1-5Hz band, θ2 in the 5.1-8Hz band, α in

the 8.1-15Hz band; feature D1 described the steepness of the curve; D2 described its

sharpness; ØF informed about the average frequency of an activity in the segment. The data

of the features describing each segment were then averaged in each class, and for each class

three additional features were extracted: t% defines the time percentage of the specific class

occurence; No gives the number of segments of a specific class; L provides the information

about the average duration of the segments of a specific class in sec. In this manner the

automatic analysis provided 312 values (8 derivations x 3 classes x 13 features) from the

five-minute-sample of the analyzed EEG signal. An example of the numeric output of the

automatic analysis is presented in Table 1.

STATISTICAL ANALYSIS

The data collected from individual infants were averaged and the number of

variables taken into account was reduced by means of factor analysis. Using the principal

component analysis, three factors – Fc1, Fc2, Fc3 – were extracted, transformed by

Varimax rotation with the Kaiser normalization and the respective factor scores were

computed. Table 2 shows the list of factors identified by the factor analysis and the list of

features represented by the specific factors; furthermore the table shows data about the

eigenvalues of factors and the percentage of variance explained by these factors.

In the first phase of the statistical analysis we tested (a) the effect of brain region,

(the activity from each brain region is represented by a symbol of the individual bipolar

derivation: Fp1-C3, …, T4-O2), (b) the effect of voltage class (low-, mid-, and high-voltage

class), (c) the effect of sleep state (quiet and active sleep), and (d) the mutual statistical

dependences of these effects on all three factors using the method of General Linear Model;

the Wilks‟ multivariate test (λ) evaluated by means of F-test served as criterion.

Subsequently using the F-test, the effect of brain region upon each factor was tested

separately, as well as the effect of voltage class, the effect of sleep state and mutual

dependences of these effects.

In the next phase of the statistical analysis, in order to determine the differences

between the individual brain regions, we evaluated the vector of the 13 EEG features in

each voltage class separately both in quiet and in active sleep. Using the General Lineal

Model method, we employed the multidimensional analysis of variance, which further

modifies the calculations of comparative tests with regard of mutual correlations between

the 13 features, so that the final tests are not affected by these correlations. Following the

initial parallel analysis of the 13 features, we compared in detail the effects of individual

brain regions for each of the 13 features using the test according to Šidák. These tests for

the individual features serve as an explanatory supplement to the basic multidimensional

tests and they illustrate which brain areas and which features participate in the topographic

differences, and the direction of these differences.

RESULTS

The effect of brain region

By evaluating the effect of brain region we were testing the presence of topographic

differentiation of EEG activity. The influence upon the factors identified by factor analysis

are shown in Table 3. It is apparent that both the entire set of factors – Fc1, Fc2, Fc3 – and

even each individual factor are highly significantly influenced by the brain region. This

means that both the factor Fc1 representing above all spectral features, and the Fc2 and Fc3

factors, which represent non-spectral features, are influenced.

The effect of voltage class and dependence between the effects of brain region and voltage

class (Table 3)

When analyzing the effect of voltage class we were testing whether the studied

features of EEG signal differ significantly in individual voltage classes. We established that

the effect of voltage class significantly influenced the entire set of factors as well as each

factor in particular. We have also found a statistically significant dependence between the

effect of brain region and the effect of voltage class for all three factors together and for

each factor in particular, which points to the fact that the effect of brain region is different in

each voltage class.

The effect of sleep state and dependence between the effects of brain region and sleep state

(Table 3)

The sleep state also significantly influenced all the analyzed factors together as well

as each factor separately. We have also proven the presence of a significant dependence

between the effect of brain region and the effect of sleep state, documenting that the brain

region effect is influenced by the sleep state, for all the three studied factors as a whole and

for factors Fc1 and Fc3.

The effect of brain region on the EEG features

The results of the comparison of measured values of the EEG features between the

individual brain areas with respect to the voltage class and to the sleep state are depicted

synoptically in Fig. 2. First we compared the data from the specific regions of the given

hemisphere to one another, so that each region was compared to the other regions of the

hemisphere (Fp1-C3 vs. C3-O1, …, C4-O2 vs. T4-O2; the activities from the studied

regions are in this case represented by the symbols of the specific derivations). Then we

compared the data from the homotopic regions of the two hemispheres to one another (Fp1-

C3 vs. Fp2-C4, …, T3-O1 vs. T4-O2). In this way we have mutually compared the activity

from the total of 12 pairs of regions altogether. In each pair of regions we were comparing

39 pairs of items (13 features x 3 voltage classes). In the end we have acquired 468 items

(12 pairs of areas x 39 pairs of items) for each sleep state, which provide the information on

the occurrence of statistically significant differences between the compared data, or lack

thereof.

While comparing data describing the activities from the individual regions of the

same hemisphere, we have found a large number of significantly different values. (a) We

have found most differences between the activities of the anterior and posterior medial

regions (Fp1,2-C3,4 vs. C3,4-O1,2). It became evident that spectral features (AV, …, D2)

and the feature No have significantly higher values in posterior regions. On the other hand

the low voltage class values of the L and t% features were significantly higher in anterior

regions. (b) The differences in activities of the anterior and posterior temporal regions

(Fp1,2-T3,4 vs. T3,4 – O1,2) were distinguished by the following fact: While the majority

of spectral features (AV, …θ2) of the high-voltage and mid-voltage class reached

significantly higher values in the anterior regions, the values of the α, D1 and D2 features in

the mid-voltage and low-voltage classes were higher in the rear. The values of non-spectral

features from both regions mostly did not differ. (c) We established sleep state dependent

differences while comparing the values from the anterior-medial and anterior-temporal

regions (Fp1,2-C3,4 vs. Fp1,2-T3,4). In quiet sleep, the values of spectral features (δ2, …,

α) were higher in the medial regions, on the contrary in active sleep spectral features

presented higher values in lateral regions. (d) The differences between activities from the

posterior-medial and posterior-lateral regions (C3,4-O1,2 vs. T3,4-O1,2) were noted for

medially localized higher values of spectral features (AV, …, α). However in active sleep

the α, D1, D2 features exhibited higher values laterally. The non-spectral features L and t%

had in the low-voltage class significantly higher values temporally, while the values of the

t% and No features in the high-voltage class were higher medially.

The comparison of data from the homotopic regions of the opposite hemispheres

exhibited only small number of statistically significant differences. Only activities from the

right and left anterior-medial regions (Fp1-C3 vs. Fp2-C4) differed significantly from each

other: Most spectral features showed higher values on the right side. Lateral differences

between the other regions were rare.

Each feature contributed to the topographic differentiation to a different degree.

Features θ1, θ2, α, and δ2 exhibited the highest occurence of significantly different entries

in quiet sleep; in turn in active sleep, these were the features δ2, t%, D1 and α. In feature

ØF, we have encountered fewest significant differences.

CONCLUSION

The main objective of the study was to establish whether the applied method is

capable of discerning topographic particularities of the neonatal EEG. The statistical

analysis proved that the effect of the brain region influences all the factors, representing the

original measured EEG variables, in a highly significant manner. In this way it

demonstrated that the applied method is adequately sensitive and that it is capable to

distinguish regional specifities of the neonatal EEG. Beside that the statistical analysis

proved that all factors are significantly influenced by both voltage class and sleep state.

Mutual statistical dependences between the effects of the brain region and voltage class and

between the effects brain region and sleep state have been found.

Paired comparison of the data acquired from each region of the individual

hemisphere exhibited substantial number of significantly different values. Our findings are

thus in accord with the outcomes of the preceding studies, which suggest topographical

differences in the values of spectral energies [6-8,10] and in the EEG complexity [17-20].

Topographic differences in EEG activity are no doubt connected to the described

morphological differences of the individual regions of the neonatal brain [23, 24], to the

established regional variances in the brain metabolism [25, 26], as well as to the identified

local differences in the maturation of brain structures [27,28]. We found that spectral

features exhibited higher values in medial derivations than in lateral ones, and at the same

time higher values in posterior than in anterior regions. The above mentioned findings

apparently testify to a more advanced functional organization in the posterior-medial

regions of the brain cortex. The analysis of non-spectral features has shown that the low-

voltage and mid-voltage segments of greater length (L) occupy greater time percentage (t%)

in the activity of the anterior-medial and posterior-lateral regions. The activity of these

regions is therefore less changeable, more rigid, and apparently contributes decisively to the

low-voltage and mid-voltage part of the tracé alternant pattern.

While comparing the data measured in the homotopic regions of the two

hemispheres, we found greater number of significantly different values only between the

activities from the anterior-central regions. Right spectral features exhibited mostly higher

values than the same features on the left. When comparing the activities from the remaining

homotopic regions of the two hemispheres, we have not found any other marked

differences. Consequently it became evident, that when using our method, the neonatal EEG

activity appears predominantly symmetrical. Other authors have come to a similar

conclusion [8, 29, 30]. One of the probable causes of the bilateral EEG symmetry are the

connections running through the corpus callosum. The described symmetry of the EEG

activity can, however, also support the idea that the functional organization of the majority

of homotopic cortical regions is not yet laterally distinguished in the neonatal period.

In the present study, we have shown that the applied automatic method is capable of

discerning the differences in the EEG signals from the different regions of the neonatal

brain. We have also proven that the topographic differences in the neonatal EEG pertain not

only spectral measures, as it is evident from the preceding computer-aided studies, but that

the topographic differences also pertain the shape and variance of the EEG signal – a fact

that has so far solicited no attention. We believe that the discriminatory capabilities of the

used method represent a promise for its application in clinical practice.

REFERENCES

1 De Weerd AW. Atlas of EEG in the first months of life. Amsterdam: Elsevier;

1995.

2 Mizrahi EM, Hrachovy RA, Kellaway P. Atlas of neonatal electro-encephalography.

3rd ed. Philadelphia: Lippinncot, Williams and Wilkins; 2004.

3 Stockard-Pope JE, Werner SS, Bickford R. G. Atlas of neonatal

electroencephalography. 2nd ed. New York: Raven Press; 1992.

4 Eiselt M, Schendel M, Witte H, Dorschel J, Cruzi-Dascalova L, D`Allest AM et al.

Quantitative analysis of discontinuous EEG in premature and full-term newborns

during quiet sleep. Electroencephalogr clin Neurophysiol, 1997;103:528-534.

5 Paul K, Krajča V, Roth Z, Melichar J, Petránek S. Quantitative topographic

differentiation of the neonatal EEG. Clin Neurophysiol, 2006;117:2050-2058.

6 Scher MS, Steppe DA, Sclabassi RJ, Banks DL. Regional differences in spectral

EEG measures between healthy term and preterm infants. Pediatr Neurol,

1997;17:218-223.

7 Thordstein M, Flisberg A, Lofgren N, Bagenholm R, Lindecrantz K, Wallin BG et

al. Spectral analysis of burst periods in EEG from healthy and post-asphyctic

neonates. Clin Neurophysiol, 2004;115:2461-2466.

8 Willekens H, Dumermuth G, Duc G, Meith D. EEG spectral power and coherence

analysis in healthy full-term neonates. Neuropediatrics, 1984;15:180-190.

9 Duffy FH, Als H, Mc Anulty GB. Infant EEG spectral coherence data during quiet

sleep; unrestricted principal components analysis-relation of factors to gestational

age, medical risk and neurobehavioral status. Clin Electroencephalogr, 2003;34:54-

69.

10 Eiselt M, Schindler J, Arnold M, Witte H, Zwiener U, Frenzl J. Functional

interactions within the newborn brain investigated by adaptive coherence analysis of

EEG. Neurophysiol Clin, 2001;31:104-114.

11 Kuks JBM, Vos JB, O`Brien MJ. EEG coherence function for normal newborns in

relation to their sleep state. Electroencephalogr clin Neurophysiol, 1988;69:295-302.

12 Parmelee AH, Akyiama Y, Schulz M, Wenner WH, Schulte JF. Analysis of the EEG

of sleeping infants. Activ Nerv Super, 1969;11:111-115.

13 Prechtl HFR, Vos JE, Akyiama Y, Casaer P. Neonatal EEG: Neonatal EEG

coherence function in relation to intrauterine growth and oestrogen levels.In: Jílek L,

Trojan S, editors. Ontogenesis of the brain, vol.2. Praha: Universita Karlova;

1974;201-210.

14 Scher MS, Jones BL, Steppe DA, Cork DL, Seltman HJ, Banks D. Functional brain

maturation in neonates as measured by EEG-sleep analysis. Clin Neurophysiol,

2003;114: 875-882.

15 Hughes JR, Kohram MH. Topographic mapping of the EEG in premature infants

and neonates. Clin Electroencephalogr, 1989;20:228-234.

16 Mandelbaum DE, Krawciw N, Assing E, Ostfeld B, Washburn D, Rosenfeld D et

al. Topographic mapping of brain potentials in the newborn infant: the estabilishment

of normal values and utility in assessing infants with neurological injury. Acta

Pediatr, 2000; 89:1104-1110.

17 de la Cruz MD, Manas S, Pereda E, Garrido JM, Lopez S, De Vera R et al.

Maturational changes in the interdependecies between cortical brain areas of

neonates during sleep. Cerebr Cortex, 2007;17:583-590.

18 Meyer-Lindenberg A. The evolution of comlexity in human brain development: an

EEG study. Electroencephalogr clin.Neurophysiol, 1996;99:405-411.

19 Pereda E, de la Cruz DM, Maňas S, Garrido JM, Lopez S, De Vera R et al.

Topography of EEG complexity in human neonates: Effect of postmenstrual age and

sleep state. Neurosci Lett, 2006;394:152-157.

20 Pereda E, Maňas S, De Vera L, Garido JM, López S, Gonzáles JJ. Non-linear

asymetric interdependencies in the electroencephalogram of healthy term neonates

during sleep. Neurosci Lett, 2003;337:101-105.

21 Krajča V, Petránek S, Patáková J,Varri A. Automatic identification of significant

graphoelements in multichannel EEG recordings by adaptive segmentation and

fuzzy clustering . Int Biomed Comput, 1991;28:71-89.

22 Paul K, Krajča V, Roth Z, Melichar J, Petránek S. Comparison of quantitative EEG

characteristics of quiet and active sleep in newborns. Sleep Med, 2003;4:543-552.

23 Conel J. The postnatal development of the human cerebral cortex. Vol.6.

Cambridge: Harvard University Press; 1939-1963.

24 Salamon G. Magnetic resonance imaging of the human brain: an anatomic atlas.

New York: Raven Press; 1990.

25 Chugani HT, Phelps ME. Maturational changes in cerebral function in infants

determined by 18FDG positron emission tomography. Science, 1986; 231: 840-843.

26 Chugani HT, Phelps ME. Imaging human brain development with positron

emission tomography. J Nucl Med, 1990;32:23-25.

27 Barchovich AJ. Normal development of the neonatal and infant brain. In:

Barchovich AJ, editor. Pediatric neuroimaging. New York: Raven press; 1990.

28 Huttenlocher PR, Dabholkar AS. Regional differences in synaptogenesis in human

cerebral cortex. J Compar Neurol, 1997;387:167-178.

29 Varner JL, Ellingson RJ, Danahy T, Nelson B. Interhemispheric amplitude

symmetry of the EEGs of normal full-term newborns. Electroencephalogr clin

Neurophysiol, 1976; 40:215-216.

30 Sterman MB, Harper RM, Havens B, Hoppenbrowers T, Mc Ginty DJ, Hodgman

JE.Quantitative analysis of infant EEG development during sleep. Electrencephalogr

clin neurophysiol, 1977;43:371-385.

ACKNOWLEDGEMENTS

This work was supported by the research program “Information Society” under Grant No.

1ET101210512 “Intelligent methods for evaluation of long-term EEG recordings” , and by

Grant IGA MZ ČR 1A8600.

Fig. 1.

100μV; 1 sec

Fp1-C3

C3 -O1

Fp1-T3

T3 -O1

Fp2-C4

C4 -O2

Fp2-T4

T4 -O2

PNG

EOG

ECG

EMG

QS AS

QS

AS

Fig. 2

AV Mm δ1 δ2 θ1 θ2 α D1 D2 ØF t% No L AV Mm δ1 δ2 θ1 θ2 α D1 D2 ØF t% No L A A P P P P ┐3 Fp1-

C3 x

C3-O1

3┌ P P

P P P P P P P P P A ┤2 2├ A P P P P P P P

P P P P P P P P P A P A

┘1 1└

P P P P P P P P A A

A A A P A P P ┐ Fp2-C4 x

C4-O2

┌ P P

P P P P A P A ┤ ├ A A A P P P P P P

P P P P P P P P A A P A

┘ └

P P P P P P P P P A P A

A A A A A ┐ Fp1-T3 x

T3-O1

┌ A

P A A P P P P ┤ ├ A A A P P P

P P P

┘ └

A A A P P

A A A A A A A ┐ Fp2-T4 x

T4-O2

┌ A

A P P P ┤ ├ A A A P P P P

P P P P A

┘ └

P P P

L M M M ┐ Fp1-C3 x

Fp1-T3

┌

L M M M M ┤ ├ L L M L L L M

M M M M

┘ └

L M L L L L L

M M M M ┐ Fp2-C4 x

Fp2-T4

┌

M M M M ┤ ├ L L M L L L L M

M M M M L

┘ └

M L L L

L M M M M M M ┐ C3-O1 x

T3-O1

┌ M M M

L M M M M L M L ┤ ├ M L M L L M M

M M M M M M L L

┘ └

M M M M M L L L L

M M M M M M M M ┐ C4-O2 x

T4-O2

┌ M M M

M M M M M L L ┤ ├ M M M L L L M M

M M M M M L M L

┘ └

M M M M M L L L L L

D D D D D D S ┐ Fp1-C3 x

Fp2-C4

┌

D ┤ ├ D D D D D S

S D

┘ └

D

D D ┐ C3-O1 x C4-O2

┌

┤ ├

┘ └

┐ Fp1-T3 x Fp2-T4

┌

S ┤ ├

┘ └

┐ T3-O1 x T4-O2

┌

S ┤ ├ S

D ┘ └

Table 1

An example of the automatic analysis output

Fp1-C3

AV Mm δ1 δ2 θ1 θ2 α D1 D2 ØF t% No L

3: 29.5 119.0 134.1 94.4 54.0 32.5 15.7 25.3 26.3 2.4 21 27 2.4

2: 17.6 73.3 80.9 50.2 33.9 22.1 11.6 19.7 21.5 2.5 31 33 2.8

1: 11.1 44.7 50.0 28.0 18.1 11.9 6.7 12.4 14.8 3.0 48 37 3.9

Numerical data obtained by the analysis of a 5–minute-period of the EEG activity from the

channel Fp1-C3 in quiet sleep. 1,2,3, voltage classes; AV,…,L, features.

Table 2

Features' representation, eigenvalues and percentage of variance of factors identified by

factor analysis

Factors Representation of features Eigenvalues % of variance

Fc1 AV,Mm,δ1,δ2,θ1,θ2,α,D1,D2, ØF 7.38 56.76

Fc2 No,t% 1.49 11.47

Fc3 L 1.31 10.08

Table 3

The effects of brain area, voltage class and sleep stat upon the factors identified by factor

analysis and the effects' interactions

Effects Brain area Volt. class Sleep state Area x Class Area x

Sleep

F p F p F p F p F p

Factors

Fc1,Fc2,Fc3 4.76 < .001 317.00 < .001 298.66 < .001 6.86 < .001 2.27 =

.001

Fc1 3.60 < .001 610.09 < .001 46.21 < .001 2.51 = .002 3.06 =

.003

Fc2 5.19 < .001 258.64 < .001 436.92 < .001 13.16 = .037 1.24 =

.274

Fc3 6.33 < .001 412.28 < .001 452.85 < .001 4.72 < .001 2.50 =

.015