Page 1
www.elsevier.com/locate/psychresns
Psychiatry Research: Neuroim
Correlation of PET and qEEG in normal subjects
Kenneth R. Alpera,b,*, E. Roy Johna,c, Jonathan Brodied,
Wilfred Gunthere, Raoul Daruwalaa, Leslie S. Prichepa,c
aBrain Research Laboratories, Department of Psychiatry, New York University Medical Center, 550 First Avenue, New York, NY 10016, USAbComprehensive Epilepsy Center, Department of Neurology, New York University Medical Center, NY, USA
cNathan S. Kline Institute for Psychiatric Research, Orangeburg, NY, USAdDepartment of Psychiatry, New York University Medical Center, New York, NY, USA
eDepartment of Psychiatry, University of Munich, Munich, Germany
Received 5 May 2005; received in revised form 15 June 2005; accepted 15 June 2005
Abstract
Positron emission tomography (PET) and quantitative electroencephalography (qEEG) were obtained in 15 normal male
subjects with eyes closed at rest. Correlations between qEEG variables and regional metabolism were examined as an approach to
investigating the metabolic and neuroanatomical basis of the generation of the EEG. Analogous to the neurometric approach to
qEEG, a normative 2-fluoro-deoxyglucose voxel data base was developed for the PET image. The PET image was transformed to
an idealized cylindrical set of coordinates to allow registration with the Talairach stereotactic atlas. PET regions of interest for the
thalamus, the left and right temporal lobes, the medial frontal cortex and the dorsolateral prefrontal cortex were defined using
Talairach coordinates and correlated to the QEEG. Salient findings included a negative correlation of thalamic metabolism to alpha
power and a positive correlation of medial frontal cortical metabolism to delta EEG power. The significance of these findings is
discussed with reference to the existing literature on the physiology of the generation of the EEG.
D 2006 Elsevier Ireland Ltd. All rights reserved.
Keywords: Human; Positron Emission Tomography (PET); Electroencephalogram (EEG); Alpha EEG power; Delta EEG power; EEG generator;
Thalamus
1. Introduction
A relationship of electroencephalographic (EEG)
activity to brain metabolism is expected because the
summed inhibitory and excitatory post-synaptic poten-
tials that are thought to account for most of the EEG are
the product of modulations of ionic conductances that
0925-4927/$ - see front matter D 2006 Elsevier Ireland Ltd. All rights rese
doi:10.1016/j.pscychresns.2005.06.008
* Corresponding author. Brain Research Laboratories, Department
of Psychiatry, New York University Medical Center, 550 First Ave-
nue, New York, NY 10016, USA. Tel.: +1 212 263 6287; fax: +1 212
263 6457.
E-mail address: [email protected] (K.R. Alper).
require metabolic energy (Pilgreen, 1995). One would
expect that there would be distinctive relationships
between regional metabolism and different EEG fre-
quency bands. However, the power in different EEG
frequency bands may be correlated with metabolism in
different brain regions. Early studies involving mea-
surement of global cerebral blood flow reported a gen-
erally positive correlation with mean EEG frequency
(Ingvar et al., 1979; Kuschinsky, 1993). However, a
simple overall correspondence between mean EEG fre-
quency and global metabolic activity might be more
characteristic of structural neurological pathology such
as primary dementia or cerebrovascular disease. For
aging 146 (2006) 271–282
rved.
Page 2
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282272
example, in two studies comparing normal controls and
demented patients (Nagata et al., 1994; Obrist et al.,
1963) and in two other studies making comparisons
between hemispheres within subjects with unilateral
infarcts (Nagata, 1989; Tolonen and Tournoux, 1981),
the amount of variance in mean EEG frequency
explained by cerebral blood flow (CBF) or metabolism
was relatively greater in the demented subjects or the
infarcted hemisphere. On the other hand, studies of
schizophrenic patients, who evidence less structural
pathology than patients with primary dementia or cere-
bral infarcts, report both negative and positive correla-
tions of positron emission tomography (PET)
metabolism to specific EEG power bandwidths depend-
ing on the anatomical region examined (Alper et al.,
1998; Gunther, 1992). Okyere et al. (1986) reported a
significant positive correlation of quantitative EEG am-
plitude with CBF in normal subjects and did not find a
significant correlation of CBF to quantitative electro-
encephalographic (qEEG) frequency. It would appear
from the limited available data that metabolic activity
may be correlated with both qEEG amplitude and
frequency in humans. Further, the relationship of me-
tabolism to qEEG power spectral composition may vary
differentially with respect to QEEG topography and
metabolic regions of interest.
A limited amount of research has been reported on
the simultaneous measurement of metabolic and EEG
data in normal, awake humans. Much of this research
has involved either correlations of EEG to metabolism
in subjacent cortex or a specific focus on the relation-
ship of EEG alpha to metabolic activity in the thalamus.
Buchsbaum et al. (1984) reported on correlations of
EEG magnitude (the square root of absolute power)
with metabolism measured utilizing PET in cortical
areas subjacent to EEG electrodes in a 16-lead array
in the left hemisphere of six normal subjects at rest with
eyes closed. Statistically significant negative correla-
tions with metabolism were found for delta and alpha
magnitude (the square root of power) in the temporal
lobe, and for alpha in the occipital lobe. Also looking at
the relationship of EEG to metabolism measured utiliz-
ing PET in cortical regions subjacent to EEG leads,
Cook et al. (1998) studied six subjects with a behav-
ioral protocol that involved sequential intervals of rest
versus motor task performance with eyes open and eyes
closed, and computed both absolute and relative EEG
power with three different types of montages: linked
ears referential, bipolar, and source-derivation. Results
differed as a function of choice of montage and absolute
versus relative power, but shared the common attribute
of generally positive correlations of metabolism to
power in the frequencies corresponding approximately
to the delta, theta and beta bands; with negative corre-
lations in the alpha band.
More recently, in a study on a combined group of 12
normal and 17 depressed subjects in which PET metab-
olism and EEG were co-registered using low frequency
electromagnetic topographic analysis (LORETA), Pizza-
galli et al. (2003) found voxelwise correlations between
normalized PET metabolism in the rostral anterior cin-
gulate gyrus and theta absolute power that were positive
in the rostral anterior cingulate cortex and middle and
superior right frontal and temporal gyri, and negative in
occipital cortex. These subjects were evaluated while at
rest with eyes closed during 50% of the uptake period,
and with eyes open during the other 50% of the uptake
period. In this study, significant voxelwise correlations
involving seven EEG bands between 1.5 and 44 Hz were
limited to the theta bandwidth. A subsequent analysis of
the 12 normal subjects found a general relationship of a
greater number of positive correlations to metabolism
with increasing frequency and negative correlations with
the low alpha (Oakes et al., 2004), consistent with earlier
work indicating a general correlation of global cerebral
blood flow or metabolism with overall EEG mean fre-
quency (Ingvar et al., 1979; Kuschinsky, 1993).
Some studies have focused on thalamic metabolic
regions of interest in studies intended to focus on the
investigation of the generation of the alpha EEG
rhythm. Larson et al. (1998) found a negative corre-
lation of thalamic metabolism measured by PET to
alpha EEG power in a combined group of 8 normal
and 19 depressed subjects. Alpha EEG power was a
single measure derived from all leads over the entire
head. As in the studies above by Pizzagalli et al.
(2003) and Oakes et al. (2004), the subjects were
evaluated while at rest with eyes closed during 50%
of the uptake period, and eyes open during the other
50% of the uptake period. In a subsequent report
involving a larger group of subjects, the negative
correlation of thalamic metabolism and EEG alpha
activity was significant only in the group of 13 normal
subjects, and not the depressed subjects. On the other
hand, Danos et al. (2001) found a positive correlation
of thalamic metabolism measured utilizing PET and
alpha EEG magnitude in 13 normal subjects during a
visual continuous performance task (CPT). In this
study the QEEG measures were derived from a 32-
lead array from which a subset of three occipital leads
were selected for the analysis. Sadato et al. (1998)
also found a positive correlation of thalamic PET
metabolism with alpha absolute power in eight normal
subjects, evaluated with eyes open during a condition
Page 3
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282 273
that combined sequential periods of rest and listening
to music. Alpha absolute power was a single measure
derived from the four posterior leads of a 12-lead
array. Goldman et al. (2002) found a positive correla-
tion of thalamic metabolism measured utilizing func-
tional magnetic resonance imaging with a single
measure of alpha absolute power averaged from a
subset of four bipolar pairs containing the occipital
electrodes.
The above studies have generally either focused on
the relationship of metabolism in cortex with voxelwise
co-registration with EEG or immediately subjacent to
EEG leads and EEG power, or the relationship of the
occipital cortical alpha EEG rhythm to a thalamic met-
abolic region of interest. The restriction of the data
analysis to a small number of variables in many of
these studies contributes statistical power, but has
resulted in the exclusion from the analysis of data
regarding non-alpha EEG bandwidths or the correlation
of EEG to non-local metabolic regions of interest. The
discarding of significant amounts of information of
potential interest itself raises statistical issues of con-
cern. Useful preliminary information regarding the neu-
rophysiological basis of the generation of the EEG may
be derived from examination of the topography of the
correlations of EEG spectral power with different met-
abolic regions of interest with all four bands of the EEG
over the full complement of the 19 standard leads of the
10–20 system.
In the present study, we report on the correlation of
qEEG power and regional metabolism in normal sub-
jects. The primary objective of this preliminary inves-
tigation is to generate images of the tomography of PET
EEG correlations in normal subjects. The metabolic
regions of interest, namely the medial frontal cortex
(MFC) (medial prefrontal, orbital, and anterior cingu-
late), the dorsolateral prefrontal cortex (DLPFC), the
thalamus (Thal), and the left and right temporal lobes
(LT and RT) were selected as metabolic regions of
interest for the prominence of their involvement in
neuropsychiatric disorders, and in addition the thalamus
was selected because of the attention focused on it in
prior studies quoted above correlating metabolism and
EEG. The intended significance of these images relates
to the understanding of the metabolic and neuroana-
tomical basis of the generation of the EEG. This study
used the statistical probability image (SPI) approach,
which involves voxel-by-voxel Z transformation of
PET data with reference to a normative data set and
allows co-registration of the PET image and the Talair-
ach coordinate system (John et al., 1994b; Talairach and
Tournoux, 1988).
2. Materials and methods
2.1. Subjects
Written informed consent was obtained from all
subjects. The index subjects used to compute these
PET–EEG relationships were 15 right-handed normal
males with a mean age of 25.2F5.3 years. The index
subjects were compared with a control group of 25
right-handed normal males that formed the voxel nor-
mative database described in John et al. (1994b). Iden-
tical selection criteria were used for both the index
subjects and the subjects from the voxel normative
database. Subjects were included who evidenced no
significant abnormality on physical examination, med-
ical history, routine blood chemistry and urine analysis,
and MRI of the head. Subjects were excluded who had
past histories of head trauma with loss of conscious-
ness, seizure, neurological or other medical conditions
known to affect the brain. Also excluded were subjects
with histories of substance abuse, psychopharmacolog-
ic treatment, psychiatric hospitalization, or evidence of
a current DSM-III-R disorder or past DSM-III-R mood
anxiety or psychotic disorder.
2.2. PET data acquisition
All subjects were scanned at the Department of
Chemistry, Brookhaven National Laboratory, on a
CTI 931 Tomograph (15 slices; spatial resolution of
6�6�6.5 mm full width at half maximum).
The subjects were positioned in the gantry accord-
ing to methods described elsewhere (Bartlett et al.,
1988), and they received a 68Ge/68Ga transmission
scan that provided attenuation correction and defined
the skull for image reconstruction and PET/MRI co-
registration. A 5–6 mCi bolus of [18F]2-fluoro-2-
deoxy-d-glucose (FDG) was prepared (Hamacher et
al., 1986) and administered (John et al., 1994a; Bar-
tlett et al., 1988) as described elsewhere. Scanning
began 35 min after FDG administration and continued
for 20 min. The subjects were scanned with their
eyes closed and ears unoccluded in a quiet, dimly
lit room.
The adequacy of counts and plasma input function
were ascertained for all scans. To minimize the effect of
region-independent global changes in glucose utiliza-
tion, relative metabolic values were calculated as the
ratio of voxel /whole brain activity (John et al., 1994b).
This reports the correlation of qEEG with relative PET
metabolism utilizing the approach to defining PET
regions of interest (ROIs) that is described below.
Page 4
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282274
2.3. EEG data acquisition
The extremely high test–retest reliability of the
qEEG power bands recorded from resting normal sub-
jects has been repeatedly confirmed (Kaye et al., 1981;
Fein et al., 1983, 1984; Kondacs and Szabo, 1999).
PET and QEEG were recorded in two separate sessions,
and EEG data were obtained within 1 week of the PET
study. The subjects were seated comfortably in a light-
attenuated room, while 20 min of eyes closed resting
EEG data were collected using a Spectrum 32 acquisi-
tion system (Cadwell Laboratories) from the 19 mono-
polar electrode sites of the International 10/20 system,
referred to linked earlobes. A differential eye channel
was used for the detection of eye movement. All elec-
trode impedances were below 5000\ohm. The EEG
amplifiers had a bandpass from 0.5 to 70 Hz (3-dB
points), with a 60-Hz notch filter. Data were sampled at
200 Hz with 12-bit resolution.
2.4. PET data analysis
The Statistical Probability Image (SPI) approach of
John et al. (1994b) was used in the analysis of PET
data. Key elements of SPI include:
1. Shape and size normalization of the PET image;
2. Z-transformation of the PET data with reference to a
normative data set;
3. Registration of normalized PET image with the
Talairach system of coordinates (Talairach and Tour-
noux, 1988).
As the first step of shape normalization of the PET
image, the edges of the image are defined on the basis
of large changes in the variance of density of voxels
(step 0). A centroid is then computed for the area within
the set of edges (Step 1). The centroid becomes the
origin for an idealized circle with radius R that circum-
scribes the slice (Step 2).
A line ri is drawn from the centroid origin to the
edge of the slice along each of 720 radii, each succes-
sively rotated from the vertical axis to an angle h. Theselines are then linearly stretched by an amount R / ri to
dilate each line to a bstandard circleQ with radius R.
Each voxel can now be identified at its location rijwhere rj indicates the jth interval along the radius riat angle hj (Step 3).
In Step 4, each individual expanded radius, R(h)i, isdivided into N sectors of Drj(hi). For each sector in the
standard circle, the ¯Drj; hijDrj; hij , and standard deviation
ADrj,hi, of the distribution of relative metabolic activity
for each sector of every slice were determined across a
group of 25 normal subjects (John et al., 1994b).
These distributions were tested for Gaussianity, and
deviations from normality corrected by appropriate
transforms. The value of each index was compared to
the control group normative data by computing the
difference between the index subject value and the
control group mean value and dividing by the control
group standard deviation (Z-transform). To reconstruct
the entire brain, the individual PET slices transformed
to standard circles, with thickness 6 mm, were stacked
to form the standard cylinder. The standard cylinder
achieves the conformed mapping of individual brains to
the standard cylindrical space, in order to facilitate
statistical comparisons by minimizing the effect of
individual anatomical variations.
To use the Talairach coordinate system to define
PET ROIs, it was necessary to bring the PET image
into registration with the standard stereotaxic atlas
(Talairach and Tournoux, 1988). The standard Talairach
stereotaxic atlas was modified by the superimposition
of a grid on each coronal section, dividing the propor-
tional scale into 10 horizontal intervals at each of 13
vertical levels. Each coronal slice was thus represented
as 120 rectangles. Each rectangle is about 8�14 mm,
or about 4� the standard deviation of localization in the
standard circle. The standard circle representing all the
slices from the whole brain can be stacked into a
standard cylinder from which transaxial, sagittal or
coronal sections can be computed. By slicing the Z-
transformed PET images, stacked into the standard
cylinder, into 16 coronal sections, about 16 slices�10
horizontal intervals�13 vertical levels=2080 ROIs can
be defined. Each of these ROIs can be thought of as
analogous to a volume element of a bMercatorQ projec-tion of the brain. Norms are computed for each bmulti-
voxel ROIQ instead of for each voxel. Regions of inter-
est subsuming multiple bmulti-voxel ROIsQ were de-
fined in Talairach coordinates.
Table 1 summarizes the Talairach coordinates of the
PET regions of interest used in the present study. The
origin of the coordinate system of Table 1 is the inter-
section of the anterior commissure–posterior commis-
sure (AC–PC) line, the vertical line traversing the
posterior margin of the anterior commissure (AC) ori-
ented perpendicular to the AC–PC line, and the midline
(Talairach and Tournoux, 1988). Distances from the
origin are expressed in centimeters. For Z coordinates,
values are negative for regions inferior to the AC–PC
line and positive for regions superior to the AC–PC
line. For Y coordinates, values are negative for regions
posterior to the AC and positive for regions anterior to
Page 5
Table 1
Talairach coordinates for PET regions of interest
Region X coordinates Y coordinates Z coordinates
Thalamus (Thal) F(0.5 to 1.7) �2.1 to �0.7 +4 to +16
F(1.0 to 2.0) �2.9 to �2.1 +4 to +12
Left temporal
(LT) or right
temporal (RT)
F(3.4 to 5.6) �4.2 to 1.8 �20 to �12F(3.4 to 6.0) �5.5 to �0.8 �12 to �1F(4.0 to 6.0) �6.2 to �1.5 �1 to +16
Medial frontal
cortex (MFC)
F(0.3 to 3.8) +5.3 to +6.6 �20 to �12F(0.3 to 1.9) +12.0 to +5.3 �24 to �14F(1.9 to 3.8) +5.3 to +6.6 �24 to�20
Dorsolateral
prefrontal
cortex
(DLPFC)
F(0 to 2.1) +5.0 to +5.6 +28 to +35
F(0.1to 1.9) +5.3 to +6.6 0 to +26
F(3.4 to 4.0) +3.6 to +5.0 +28 to +35
F(4.0 to 5.4) +3.7 to +5.3 +4 to +24
F(4.8 to 5.9) 0 to +1.8 +18 to +32
F(4.6 to 5.2) +2.3 to +4.1 +6 to +24
F(5.0 to 5.3) 0 to +1.8 +35 to +40
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282 275
the AC. For X coordinates, values to the left of the
midline are positive and those to the right are negative.
The ROIs thalamus (Thal), medial frontal cortex
(MFC), and dorsolateral prefrontal cortex (DLPFC)
are each composed of two symmetrical left and right
halves with identical Y and Z coordinates. Left temporal
(LT) and right temporal lobe (RT) are each considered
separately and are the mirror images of one another.
In selecting coordinates to define the ROIs, an
attempt was made to circumscribe rectangular volumes
entirely within the structure of interest and to mini-
mize the influence of white matter. For the LT and RT,
the coordinates subsumed Brodmann areas 21 through
24, 28, 37, 38, 41 and 42. MFC was defined to
subsume the medial prefrontal and orbital cortex, and
anterior cingulate, and consisted of the medial portion
of area 10, and areas 11, 12, 32 and 33. DLPFC
consisted of the lateral portion of area 10, and areas
9, 44 and 46.
2.5. QEEG data analysis
A computerized artifact-detection algorithm and vi-
sual inspection were used to obtain 48 epochs (2.5 s
each, for a total of 2 min) of artifact-free data from 20
min of continuous raw EEG for quantitative analysis.
Power spectral analysis was performed using Fast Four-
ier Transform (FFT). For each of the 19 monopolar
derivations, absolute and relative (%) power, and
mean frequency, were computed for the delta (1.5–3.5
Hz), theta (3.5–7.5 Hz), alpha (7.5 to 12.5 Hz), and beta
(12.5–25 Hz) frequency bands. Using Neurometrics,
quantitative electrophysiologic features are log-trans-
formed to obtain Gaussianity, age-regressed, and Z-
transformed relative to population norms. The impor-
tance of each of these steps in enhancing the sensitivity
and specificity of electrophysiological data has been
discussed in detail elsewhere (John et al., 1987,
1988). Independent replications have confirmed the
Neurometric qEEG norms (Matousek and Petersen,
1973; Ahn et al., 1980; John et al., 1987, 1989; Gasser
et al., 1982; Jonkman et al., 1985; Alvarez et al., 1987;
Harmony et al., 1987; Diaz de Leon et al., 1988;
Veldhuizen et al., 1993; Duffy et al., 1993, 1995),
and provide evidence of high test retest reliability
(Kaye et al., 1981; Fein et al., 1983, 1984; Kondacs
and Szabo, 1999).
In addition to the univariate spectral features, multi-
variate regional composite features are also computed.
These features represent multivariate abnormalities for
specific spectral features across a set of topographic
regions. In the context of this study, these composites
were considered to be similar to the ROIs used in the
PET data analysis.
2.5.1. Data reduction
Data reduction of the qEEG was required to reduce
the large number of extracted features for the purpose
of input to the canonical correlation analyses. To
achieve the desired reduction and feature selection,
Pearson r correlations were computed between each
ROI and the qEEG variables, separately for each mea-
sure set. These correlations were topographically dis-
played and used to aid in the selection of qEEG bROIsQ,those regions with maximum correlations to the PET
ROI, and are shown in Figs. 1 and 2. This approach
follows that described by others (Weiner and Dunn,
1966; Oken and Chiappa, 1986), using a t-test or
ANOVA model where variables are selected which
maximize adjusted multiple correlation coefficients be-
tween qEEG and dependent variables, minimizing the
residual sum of squares (RSS).
2.6. Correlation of PET and EEG
PET ROIs were defined by Talairach coordinates
that are presented in Table 1. Relative PET metabolism
for each of the above ROIs was correlated with qEEG
absolute and relative power and mean frequency. The
Pearson r correlations computed between each ROI and
the qEEG variables are displayed in topographic maps
in Figs. 1 and 2. The color-coded scale goes from a
correlation of F0.5 through zero, where for N =15
(df =13), at P=0.05, r =0.51.
The SAS/STATR CANCORR procedure was used
to compute canonical correlations. With this method,
the significance of the relationship between sets of
Page 6
Fig. 1. Correlations of PET metabolism with qEEG for delta, theta, alpha and beta frequencies (successive columns in each panel) for absolute
power (top row each panel), relative power (middle row each panel) and mean frequency (bottom row each panel) for different ROIs. Each panel is
the color-coded topographic map of the correlation values for: Medial Frontal Cortex (top left panel), Left Temporal Cortex (bottom left panel),
Dorsolateral Prefrontal Cortex (top right panel) and Right Temporal Cortex (bottom right panel) and Thalamus (bottom left panel). The color scale
for these images go from �1 (blue/light green), through zero (black) to +1 (red/yellow), for N =15 (df =13), p =0.05, r =0.51.
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282276
qEEG variables and each PET ROI was computed.
CANCORR uses an F approximation in its calculation.
This method can be considered an extension of multiple
regression and correlation analysis, where sets of vari-
ables can be used. Using CANCORR, the linear com-
binations of the sets of qEEG variables that maximize
the correlation with the PET ROI are sought.
3. Results
Maps of the correlations of each of the PET ROIs
with EEG absolute and relative power and mean fre-
quency are presented in Figs. 1 and 2. Each of the panels
in these figures depicts a single PET ROI correlating
with absolute and relative power and mean frequency
(successive rows in each panel) for the delta, theta, alpha
and beta EEG bandwidths (successive columns in each
panel). In general, it appears that correlations of PET
metabolism with qEEG show different patterns that may
be global or localized to either subjacent or distant leads.
The top panels of Fig. 1 depict the qEEG correla-
tions with PET MFC and DLPFC ROIs (left and right
panel, respectively). MFC activity is positively corre-
lated with delta power maximally at Fz and delta mean
frequency at O1. The topography of correlations for
MFC and DLPFC appears to differ. In contrast to MFC,
the strongest correlations with DLPFC activity are neg-
ative and occur for absolute power in the alpha and beta
Page 7
Table 2
PET ROI, canonical correlation with qEEG variables, Wilk’s Lambda
F and probability obtained in these analyses
PET ROI CAN CORR Wilk’s Lambda Approx F PrNF
DPFC 0.8350 0.3038 5.7295 0.012
THAL 0.9205 0.1125 5.9182 0.022
AMFC 0.6375 0.5939 4.1029 0.044
LTEMP 0.9759 0.0476 8.0103 0.029
RTEMP 0.7551 0.4363 7.7531 0.007
Fig. 2. Correlation of PET metabolism with qEEG as in Fig. 1 above, for the Thalamus. Color coding as described for Fig. 1.
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282 277
bandwidths in parietal–occipital leads. There is a pos-
itive correlation with relative delta power in F7.
The bottom panels of this figure map the correlations
of qEEG and PET LT and RT ROIs (left and right panel,
respectively). Possibly the most striking aspect of these
two sets of images is the marked difference of the
pattern of correlation of qEEG with the right versus
the left side. RT metabolic activity is correlated nega-
tively with beta absolute power frontally and delta
absolute power centrally. Positive correlations of RT
metabolism are seen with theta mean frequency in the
left temporal and central leads, and with beta absolute
power frontally. The LT region differs considerably
from the RT, with a negative relationship with theta
absolute and relative power and mean frequency tend-
ing to be maximal in right posterior temporal and
occipital leads, and generally weaker correlations than
those seen in RT.
Fig. 2 maps the correlations of qEEG and the PET
THAL ROI. It indicates a global negative relationship
between thalamic metabolism and alpha absolute and
relative power and beta mean frequency, as well as a
positive correlation with theta and beta relative power
in frontal leads. Thalamic metabolism is also correlated
with delta absolute and relative power at F7.
Using qEEG variables selected from the correlation
analyses described above, the canonical correlations
between qEEG ROIs and PET ROIs were computed.
As can be seen in Table 2, there were significant
canonical correlations for all PET ROIs. For the
DLPFC, the qEEG variables contributing most to the
significant canonical correlations were as follows: ab-
solute power in alpha for multivariate posterior regions
and in beta for posterior temporal and multivariate
anterior regions. For the THAL, the qEEG variables
contributing most to the significant canonical correla-
tion were relative power in alpha for all frontal regions
and the multivariate for posterior regions. For the
AMFC, the qEEG variables contributing most to the
significant canonical correlations were as follows: the
multivariates for anterior regions and posterior regions
in the delta band. For the LT, the qEEG variables
contributing most to the significant canonical correla-
tions were as follows: absolute power in theta in the
central and posterior regions and the multivariate ante-
rior regions. For the RTEMP, the qEEG variables con-
tributing most to the significant canonical correlations
were as follows: absolute power in beta in the multi-
variate anterior regions and mean frequency in theta in
the central regions.
4. Discussion
The most significant findings in this study appear
to include (1) a negative correlation of thalamic me-
,
Page 8
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282278
tabolism to alpha power, (2) a positive correlation of
medial frontal cortical metabolism to delta EEG abso-
lute power and beta mean frequency, (3) generally
negative correlations of DLPFC to absolute power in
all leads except for a positive correlation of delta
absolute and relative power at F7, and (4) generally
different topographic patterns of correlations of me-
tabolism to EEG in the right versus left temporal
lobes, RT correlating positively with prefrontal beta
and beta mean frequency, and LT negatively correlat-
ing with theta in posterior leads. The correlations of
PET metabolism to EEG are both global and relatively
localized and not confined to subjacent leads. While
this is a preliminary study involving a relatively small
number of subjects, previous studies using this SPI
approach (John et al., 1994b) showed significant sen-
sitivity and specificity when applied to a comparably
small clinical sample.
In addition to the relatively small number of sub-
jects, the non-simultaneity of the PET and EEG eva-
luations, which were conducted up to a week apart, is
a factor that might decrease the significance of the
PET–EEG correlations was reported here. However,
evidence for a high degree of test–retest reproducibil-
ity has been reported for both PET (Bartlett et al.,
1988) and qEEG (John et al., 1983; Kaye et al., 1981;
Fein et al., 1983, 1984; Kondacs and Szabo, 1999),
and specifically between EEG obtained in the EEG
laboratory and during the PET scanning procedure
(Alper et al., 1998).
The choice of EEG electrode montage has been
suggested to affect the robustness of PET–EEG correla-
tions. In a study on six normal subjects that compared
the use of referential, bipolar, and source derivation
montages, the source derivation produced higher corre-
lations between PET and qEEG than the referential or
bipolar montages (Cook et al., 1998). However, much
of the research on qEEG correlations with metabolism
in normal subjects has used a referential montage
(Buchsbaum et al., 1982; Danos et al., 2001; Lindgren
et al., 1999). The monopolar linked-ears montage is the
convention in the Neurometric methodology, which
offered the methodological and statistical advantages
of a large normative database and the ability to compute
QEEG measures as Z scores referenced to normative
values. A referential montage is also most suitable for
the purpose of generating maps of PET ROIs correlated
with a set of qEEG variables that most fully samples the
topography spanned by the 19 channel electrode loca-
tions, rather than limiting the analysis to pairing EEG
leads with immediately subjacent cortical metabolic
regions.
The scan for Thal (Fig. 2) and MFC (Fig. 1), seem
statistically and physiologically plausible, and indi-
cates a negative relationship between thalamic metab-
olism and EEG alpha absolute and relative power,
consistent with a similar negative relationship of
EEG alpha power to thalamic metabolism that has
been reported elsewhere in the literature (Larson et
al., 1998; Lindgren et al., 1999). A correlation of
thalamic metabolism with alpha might be expected
in view of the evidence that the thalamus is important
in the generation of alpha activity (Nunez, 1995a;
Lopes da Silva, 1991; Steriade et al., 1990). The
fact that the correlation of thalamic metabolism with
alpha power is negative is consistent with a model of
generation of the alpha EEG rhythm that emphasizes
the dynamic interaction of cortex and thalamus (Llinas
and Ribary, 2001; John, 2002; Nunez, 1995a). This
view differs distinctly from the idea of the thalamus as
a bpacemakerQ of the awake alpha rhythm, and in-
stead, involves the selective activation of different
resonant modes or bEigenstatesQ of cerebral cortex.
The frequency at which the thalamus activates the
cortex is strongly dependent on the state of membrane
polarization of the reticular nucleus of the thalamus
(RNT) (Nunez, 1995b; Silberstein, 1995); however,
the frequency with which the cortex becomes engaged
in the resulting thalamocortical oscillation also
depends on the intrinsic oscillatory modes of the
cortex (Silberstein, 1995). The rhythmic cortical mode
then becomes entrained on the thalamus reflecting
thalamic bfrequency plasticity,Q resulting in matching
of the resonant frequencies of the thalamus and cortex
in a stable and energetically conservative oscillatory
state.
According to this model, resonance of thalamus and
cortex in the alpha frequency range corresponds to a
state of relative hyperpolarization of thalamic relay
cells by RNT and decreased transfer of signal from
the periphery through thalamus to cortex. Under these
conditions, a negative correlation between thalamic
metabolism and alpha activity might be expected due
to a lower energy requirement of the thalamus during
generation of the global cortical alpha rhythm with
decreased throughput to cortex. Conversely, the rela-
tively higher energetic requirement of a depolarized
bias of thalamic relay cells by the RNT, with increased
transfer of signal via specific thalamic afferents to
cortex , would be expected to correspond to higher
frequency local cortical resonances and diminished
alpha.
This model can be reconciled with the fMRI study
of Goldman et al. (2002), which reported a positive
Page 9
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282 279
correlation between alpha power and thalamic metab-
olism. When metabolism is measured with PET in
subjects at rest with eyes closed as in the present
study, the temporally dominant feature during the
continuous radiolabeled glucose uptake period would
presumably be the energetically conservative thalamo-
cortical oscillatory state, in which lower thalamic
metabolism would be expected to correspond to great-
er alpha power. Functional magnetic resonance imag-
ing, in contrast to PET, does not measure the baseline
energetically conservative thalamocortical oscillatory
state corresponding to the alpha rhythm itself, but
instead measures the changes in cerebral blood flow
corresponding to the modulation of the alpha rhythm.
Some studies have reported a positive correlation
between thalamic metabolism measured utilizing
PET and alpha EEG power (Danos et al., 2001).
This may be due to the fact that subjects in these
studies were presented with sensory stimuli such as
the visual continuous performance task (Danos et al.,
2001) or music, as opposed to the quiet resting
condition of the studies, including the present one,
in which a negative relationship of thalamic metabo-
lism was observed (Lindgren et al., 1999; Larson,
1998).
Fig. 1 shows a positive correlation between MFC
metabolism and delta EEG activity. We have previ-
ously reported a similar positive correlation between
delta EEG power and frontal metabolic activity mea-
sured by PET in a small group of schizophrenic sub-
jects (Alper et al., 1998). Pizzagalli et al. (2003)
reported a positive correlation between theta and me-
tabolism in the anterior cingulate and temporal corti-
ces, without significant correlations to delta; however,
that study differed from the present one in a number
of respects including the definition of the delta and
theta bandwidths and regions of interest. Although
there is a tendency in the neurological literature to
ascribe only pathological significance to delta EEG
activity, there is strong evidence for delta as a corre-
late of normal mental activity in awake humans. In
multiply replicated EEG norms, delta activity accounts
for more than 20% of the raw total EEG power in
anterior leads (Matousek and Petersen, 1973; Gasser et
al., 1982; Jonkman et al., 1985; Alvarez et al., 1987;
Harmony et al., 1987; John et al., 1987). Studies using
dipole modeling have located the major delta genera-
tor deep in prefrontal cortex (Michel et al., 1992,
1993). A variety of behavioral studies of delta activity
in normal awake humans show a positive relationship
of delta activity to mental effort. Increased frontal
delta has been observed in normal subjects performing
calculations (Fernandez et al., 1995; Harmony et al.,
1996, 2004), delayed match from sample (John et al.,
1996; Harmony et al., 2004), word and figure catego-
rization (Harmony et al., 2001), reaction time tasks
(Van Dijk et al., 1992), abstract thought (Michel et al.,
1993) and an omitted stimulus paradigm (Basar-Ero-
glu et al., 1992), and to covary positively with P300
amplitude (Schurmann et al., 2001; Basar et al., 1984;
Intriligator and Polich, 1994). Fernandez et al. (1995)
hypothesize that normal physiologic delta might be a
functional correlate of the gating of peripheral input in
order to facilitate binner concentration,Q or the alloca-
tion of attention to an internally represented task.
Delta activity can also be viewed as a global rhythm
of association cortex that serves the purpose of inte-
gration by synchronizing coherent activity and phase
coupling across widely spatially distributed neural
assemblies (Nunez, 1995a).
An important issue that should be addressed in
future work correlating PET and EEG is the question
of the underlying dimensionality of the PET and EEG
measure sets. In the present study, the dimensionality
of the PET measure set was reduced by defining a
priori anatomical regions of interest. The EEG mea-
sure set in this study was defined by the descriptors
consisting of the leads of the 10–20 system and the
traditional four EEG bandwidths. The dimensionality
of the data sets can also be defined empirically by
factor analysis of either the PET or EEG data sets. The
general agreement between multiple laboratories re-
garding the results of factor analysis of PET data
(Volkow et al., 1986; Schroeder et al., 1994; Zuendorf
et al., 2003) is evidence for a replicable underlying set
of dimensions of the PET metabolic image. There are
multiple replications of small sets of factors resulting
from principal components analysis of EEG data
(Andresen, 1993; Arruda et al., 1996). In addition to
data reduction, an appreciation of the factor structure
of PET and EEG might aid the understanding of the
metabolic basis of the neurophysiologic generation
of the EEG, and might potentially add statistical
power in detecting abnormality or change associated
with neuropsychiatric disorders or psychopharmaco-
logical effects.
In summary, this preliminary study found correla-
tions between regional metabolism and EEG spectral
power. The findings of a negative correlation be-
tween thalamic metabolism and alpha power, and a
positive correlation of frontal cortical metabolism and
delta power appear plausible given the present un-
derstanding of the neurophysiological basis of these
two EEG rhythms.
Page 10
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282280
References
Ahn, H., Prichep, L.S., John, E.R., Baird, H., Trepetin, M., Kaye, H.,
1980. Developmental equations reflect brain dysfunction. Science
210, 1259–1262.
Alper, K., Gunther, W., Prichep, L.S., John, E.R., Brodie, J., 1998.
Correlation of QEEG with PET in schizophrenia. Neuropsycho-
biology 38, 50–56.
Alvarez, A., Pascual, R., Valdes, P., 1987. EEG development equa-
tions confirmed for Cuban school children. Electroencephalogra-
phy and Clinical Neurophysiology 67, 330–332.
Andresen, B., 1993. Multivariate statistical methods and their capa-
bility to demarcate psychophysiologically and neurophysiologi-
cally sound frequency components of human scalp EEG. In:
Zschocke, G., Speckmann, G. (Eds.), Basic Mechanisms of the
EEG. Birkhauser, Berlin, pp. 317–352.
Arruda, J.E., Weiler, M.D., Valentino, D., Willis, W.G., Rossi, J.S.,
Stern, R.A., Gold, S.M., Costa, L., 1996. A guide for applying
principal components analysis and confirmatory factor analysis to
quantitative electroencephalogram data. International Journal of
Psychophysiology 23, 63–81.
Bartlett, E.J., Brodie, J.D., Wolf, A.P., Chrisman, D.R., Laska, E.,
Meissner, M., 1988. Reproducibility of cerebral glucose metabolic
measurements in resting human subjects. Journal of Cerebral
Blood Flow and Metabolism 8, 502–512.
Basar, E., Basar-Eroglu, C., Rosen, B., Schutt, A., 1984. A new
approach to endogenous event related potentials in man: relation
between EEG and P300 wave. International Journal of Neurosci-
ence 26, 161–180.
Basar-Eroglu, C., Basar, E., Demiralp, T., Schurmann, M., 1992.
P300-response: possible psychophysiological correlates in delta
and theta frequency channels. A review. International Journal of
Psychophysiology 13, 161–179.
Buchsbaum, M.S., Ingvar, D.H., Kessler, R., 1982. Cerebral gluco-
graphy with positron emission tomography. Archives of General
Psychiatry 39, 251–259.
Buchsbaum, M.S., DeLisi, L.E., Holcomb, H.H., Ingvar, D.H., Kess-
ler, R., 1984. Anteroposterior gradients in cerebral glucose use in
schizophrenia and affective disorders. Archives of General Psy-
chiatry 41, 1159–1166.
Cook, J.A., O’Hara, R., Uijtdebaage, S.H.J., Mandelkern, M.,
Leuchter, A.F., 1998. Assessing the accuracy of topo-
graphic EEG mapping for determining local brain func-
tion. Electroencephalography and Clinical Neurophysiology 107,
408–414.
Danos, P., Guich, S.M., Abel, L., Buchsbaum, M.S., 2001. EEG alpha
rhythm and glucose metabolic rate in the thalamus in schizophre-
nia. Neuropsychobiology 43, 265–272.
Diaz de Leon, A.E., Harmony, T., Marosi, E., Becker, J., Alvarez, A.,
1988. Effect of different factors on EEG spectral parameters.
International Journal of Neuroscience 43, 123–131.
Duffy, F.H., Albert, M.S., McAnulty, G.B., 1993. The pattern of age-
related differences in electrophysiological activity of healthy sub-
jects. Neurobiology of Aging 14, 73–74.
Duffy, F.H., Jones, K., McAnulty, G.B., Albert, M., 1995.
Spectral coherence in normal adults: unrestricted principal
components analysis. Clinical EEG and Electroencephalography
26, 30–46.
Fein, G., Galin, D., Johnstone, J., Yingling, C.D., Marcus,
M., Kiersch, M.E., 1983. EEG power spectrum in normal
and dyslexic children: 1. Reliability during passive condi-
tions. Electroencephalography and Clinical Neurophysiology 55,
399–405.
Fein, G., Galin, D., Yingling, C.D., Johnstone, J., Nelson, M.A.,
1984. EEG spectra in 9–13 year old boys are stable over 1–3
years. Electroencephalography and Clinical Neurophysiology 58,
517–518.
Fernandez, T., Harmony, T., Rodriguez, M., Bernal, J., Silva, J.,
Reyes, A., Marosi, E., 1995. EEG activation patterns during the
performance of tasks involving different components of mental
calculation. Electroencephalography and Clinical Neurophysiolo-
gy 94, 175–182.
Gasser, T., Bacher, P., Mochs, J., 1982. Transformation towards the
normal distribution of broad band spectral parameters of the
EEG. Electroencephalography and Clinical Neurophysiology 53,
119–124.
Goldman, R.I., Stern, J.M., Engel, J., Cohen, M.S., 2002. Simulta-
neous EEG and fMRI of the alpha rhythm. NeuroReport 13,
2487–2492.
Gunther, W., 1992. MRI–SPECT and PET–EEG findings on brain
dysfunction in schizophrenia. Progress in Neuro-psychopharma-
cology & Biological Psychiatry 16, 445–462.
Hamacher, K., Coenen, H.H., Stocklin, G., 1986. Efficient stereospe-
cific synthesis of no-carrier-added 2-[18F]-fluoro-2-deoxy-d-glu-
cose using aminopoly ether supported nucleophilic substitution.
Journal of Nuclear Medicine 27, 235–238.
Harmony, T., Alvarez, A., Pascual-Marqui, R., Ramos, A., Marosi, E.,
Diaz de Leon, A., Valdes, P., Becker, J., 1987. EEG maturation of
children with different economic and psychosocial characteristics.
International Journal of Neuroscience 31, 103–113.
Harmony, T., Fernandez, T., Silva, J., Bernal, J., Diaz-Comas, L.,
Reyes, A., Marosi, E., Rodriquez, M., 1996. EEG delta activity:
an indicator of attention to internal processing during performance
of mental tasks. International Journal of Psychophysiology 24,
161–171.
Harmony, T., Fernandez, T., Fernandez-Bouzas, A., Silva-Pereyra, J.,
Bosch, J., Diaz-Comas, L., Galan, L., 2001. EEG changes during
word and figure categorization. Clinical Neurophysiology 112,
1486–1498.
Harmony, T., Fernandez, T., Gersenowies, J., Galan, L., Fernandez-
Bouzas, A., Aubert, A., Diaz-Comas, L., 2004. Specific EEG
frequencies signal general common cognitive processes as well
as specific task processes in man. International Journal of Psy-
chophysiology 53, 207–216.
Ingvar, D.H., Rosen, I., Johannesson, G., 1979. EEG related to
cerebral metabolism and blood flow. Pharmakopsychiatrie,
Neuro-Psychopharmakologie 12, 200–209.
Intriligator, J., Polich, J., 1994. On the relationship between back-
ground EEG and P300 event-related potential. Biological Psychi-
atry 37, 207–218.
John, E.R., 2002. The neurophysics of consciousness. Brain Research
Reviews 39, 1–28.
John, E.R., Prichep, L.S., Ahn, H., Easton, P., Fridman, J., Kaye, H.,
1983. Neurometric evaluation of cognitive dysfunctions and neu-
rological disorders in children. Progress in Neurobiology 21,
239–290.
John, E.R., Prichep, L.S., Easton, P., 1987. Normative data banks and
neurometrics, basic concepts, methods and results of norm con-
struction. In: Gevins, A.S., Remond, A. (Eds.), Handbook of
Electroencephalography and Clinical Neurophysiology, vol. I.
Elsevier, New York, pp. 449–495.
John, E.R., Prichep, L.S., Friedman, J., Essig-Peppard, T., 1988.
Neurometric classification of patients with different psychiatric
Page 11
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282 281
disorders. In: Samson-Dollfus, D. (Ed.), Statistics and Topography
in Quantitative EEG. Elsevier, New York, pp. 88–95.
John, E.R., Prichep, L.S., Ahn, H., Kaye, H., Brown, D., Easton, P.,
Karmel, Z., Toro, A., Thatcher, R., 1989. Neurometric Evaluation
of Brain Function in Normal and Learning Disabled Children.
University of Michigan Press, Ann Arbor, MI.
John, E.R., Prichep, L.S., Alper, K.R., Mas, F., Cancro, R., Easton, P.,
Sverdlov, L., 1994. Quantitative electrophysiological characteris-
tics and subtyping of schizophrenia. Biological Psychiatry 36,
801–826.
John, E.R., Zhang, Z., Brodie, J.D., Prichep, L.S., 1994. Statistical
probability mapping of brain function and structure. In: Thatcher,
R.W., Hallett, M., Zeffiro, T., John, E.R., Huerta, M. (Eds.),
Functional Neuroimaging: Technical Foundations. Academic
Press, New York, pp. 137–143.
John, E.R., Easton, P., Isenhart, R., Allen, P., Gulyashar, A., 1996.
Electrophysiological analysis of the registration, storage and re-
trieval of information in delayed matching from samples. Inter-
national Journal of Psychophysiology 24, 127–144.
Jonkman, E.J., Poortvliet, D.C.J., Veering, M.M., DeWeerd, A.W.,
John, E.R., 1985. The use of neurometrics in the study of patients
with cerebral ischemia. Electroencephalography and Clinical Neu-
rophysiology 61, 333–341.
Kaye, H., John, E.R., Ahn, H., Prichep, L.S., 1981. Neurometric
evaluation of learning disabled children. International Journal of
Neuroscience 13, 15–25.
Kondacs, A., Szabo, M., 1999. Long-term intra-individual variability
of the background EEG in normals. Clinical Neurophysiology
110, 1708–1716.
Kuschinsky, W., 1993. Is the EEG correlated with the brain metab-
olism and cerebral blood flow? Basic Mechanisms of the EEG,
pp. 109–119.
Larson, C.L., Davidson, R.J., Abercrombie, H.C., Ward, R.T., Schae-
fer, S.M., Jackson, D.C., Holden, J.E., Perlman, S.B., 1998.
Relations between PET-derived measures of thalamic glucose
and EEG alpha power. Psychophysiology 35, 162–169.
Lindgren, K.A., Larson, C.L., Schaefer, S.M., Abercrombie, H.C.,
Ward, R.T., Holden, J.E., Perlman, S.B., Benca, R.M., Davidson,
R.J., 1999. Thalamic metabolic rate predicts EEG alpha power in
healthy control subjects but not in depressed patients. Biological
Psychiatry 45, 943–952.
Llinas, R., Ribary, U., 2001. Consciousness and the brain. The
thalamocortical discharge in health and disease. Annals of the
New York Academy of Sciences 929, 166–175.
Lopes da Silva, F., 1991. Neural mechanisms underlying brain waves:
from neural membranes to networks. Electroencephalography and
Clinical Neurophysiology 79, 81–93.
Matousek, M., Petersen, I., 1973. Norms for the EEG. In: Kellaway,
P., Petersen, I. (Eds.), Automation of Clinical Electroencephalog-
raphy. Raven Press, New York, pp. 75–102.
Michel, C.M., Lehmann, D., Henggeler, B., Brandeis, D., 1992.
Localization of the sources of EEG delta, theta, and beta frequen-
cy bands using the FFT dipole approximation. Electroencepha-
lography and Clinical Neurophysiology 82, 38–44.
Michel, C.M., Henggeler, B., Brandeis, D., Lehmann, D., 1993.
Localization of sources of brain alpha/theta/delta activity and
the influence of the mode of spontaneous mentation. Physiolog-
ical Measurement 14, 21–26.
Nagata, K., 1989. Topographic EEG mapping in cerebrovascular
disease. Brain Topography 2, 119–128.
Nagata, K., Yokoyama, E., Hirata, Y., Hatazawa, J., 1994. Electroen-
cephalographic correlates of cortical blood flow and metabolism
in normal elderly subjects. The 8th IPEG Symposium: Abstracts,
vol. 20.
Nunez, P.L., 1995. Experimental connections between EEG data and
the global wave theory. In: Nunez, P.L. (Ed.), Neocortical Dy-
namics and Human EEG Rhythms. Oxford University Press,
London, pp. 534–590.
Nunez, P.L., 1995. Mind, brain and electroencephalography. In:
Nunez, P.L. (Ed.), Neocortical Dynamics and Human EEG
Rhythms. Oxford University Press, London, pp. 133–194.
Oakes, T.R., Pizzagalli, D., Hendrick, A.M., Horras, K.A., Larson,
C.L., Abercrombie, H.C., Schaefer, S.M., Koger, J.V., Davidson,
R.J., 2004. Functional coupling of simultaneous electrical and
metabolic activity in the human brain. Human Brain Mapping
21, 257–270.
Obrist, W.D., Sokoloff, L., Lassen, N.A., Lane, N.H., Butler, R.N.,
Feinberg, I., 1963. Relation of EEG to cerebral blood flow and
metabolism in old age. Electroencephalography and Clinical Neu-
rophysiology 15, 610–619.
Oken, B.S., Chiappa, K.H., 1986. Statistical issues concerning com-
puterized analysis of brainwave topography. Annals of Neurology
19, 493–494.
Okyere, J.G., Ktonas, P.Y., Meyer, J.S., 1986. Quantification of the
alpha EEG modulation and its relation to cerebral blood flow.
IEEE Transactions on Biomedical Engineering 33, 690–696.
Pilgreen, K.L., 1995. Physiologic, medical and cognitive correlates of
electroencephalography. Neocortical Dynamics and Human EEG.
, pp. 195–248.
Pizzagalli, D., Oakes, T.R., Davidson, R.J., 2003. Coupling of theta
activity and glucose metabolism in the human rostral anterior
cingulate. An EEG/PET study of normal and depressed subjects.
Psychophysiology 40, 939–949.
Sadato, N., Nakamura, S., Oohashi, T., Nishina, E., Fuwamoto, Y.,
Waki, A., 1998. Neural networks for generation and suppression
of alpha rhythm: a PET study. NeuroReport 9, 893–897.
Schroeder, J., Buchsbaum, M., Siegel, B.V., Geider, F.J., Haier, R.J.,
Lohr, J., Wu, J., Potkin, S.G., 1994. Patterns of cortical activity
in schizophrenia. Psychological Medicine 24, 947–955.
Schurmann, M., Basar-Eroglu, C., Kolev, V., Basar, E., 2001. Delta
responses and cognitive processing: single-trial evaluations of
human visual P300. International Journal of Psychophysiology
39, 229–239.
Silberstein, R.B., 1995. Neuromodulation of neocortical dynamics. In:
Nunez, P.L., Thierry, A.M., Glowinski, P.S., Goldman-Rakic, P.S.
(Eds.), Neocortical Dynamics and Human EEG Rhythms. Oxford
University Press, London, pp. 591–627.
Steriade, M., Gloor, P., Llinas, R.R., Lopes Da Silva, F., Mesu-
lam, M.M., 1990. Basic mechanisms of cerebral rhythmic
activities. Electroencephalography and Clinical Neurophysiology
76, 481–508.
Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the
Human Brain. Thieme Medical Publishers, New York.
Tolonen, U., Tournoux, P., 1981. Comparison of quantitative EEG
parameters from four different analysis techniques in evalua-
tion of relationships between EEG and CBF in brain infarc-
tion. Electroencephalography and Clinical Neurophysiology 51,
177–185.
Van Dijk, J.G., Caekebeke, V.J.F., Zwinderman, A.H., 1992.
Background EEG reactivity in auditory event-related poten-
tials. Electroencephalography and Clinical Neurophysiology 83,
44–51.
Veldhuizen, R.J., Jonkman, E.J., Poortvliet, J.D.C., 1993. Sex differ-
ences in age regression parameters of healthy adults—normative
Page 12
K.R. Alper et al. / Psychiatry Research: Neuroimaging 146 (2006) 271–282282
data. Electroencephalography and Clinical Neurophysiology 86,
377–384.
Volkow, N.D., Brodie, J.D., Wolf, A.P., 1986. Brain metabolism in
predominantly never-treated schizophrenics before and after acute
neuroleptic administration. Journal of Neurology, Neurosurgery
and Psychiatry 49, 1199–1201.
Weiner, J.M., Dunn, O.J., 1966. Elimination of variates in discrimi-
nation problems. Biometrics 22, 268–275.
Zuendorf, G., Kerrouche, N., Herholz, K., Baron, J.C., 2003.
Efficient principal component analysis for multivariate 3D
voxel-based mapping of brain functional imaging data sets as
applied to FDG-PET and normal aging. Human Brain Mapping
18, 13–21.