EEG source analysis of chronic fatigue syndrome Psychiatry

Psychiatry Research: Neuroimaging 181 (2010) 155–164
Contents lists available at ScienceDirect
Psychiatry Research: Neuroimaging
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p s yc h r e s n s
EEG source analysis of chronic fatigue syndrome
Pierre Flor-Henrya,⁎, John C. Linda, Zoltan J. Kolesb
a
Alberta Hospital Edmonton, Box 307, Edmonton, Edmonton, Alberta, Canada T5J 2J7
Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta, Canada
b
a r t i c l e
i n f o
Article history:
Received 27 October 2008
Received in revised form 27 August 2009
Accepted 16 October 2009
Keywords:
Chronic fatigue syndrome
Spectral analysis
EEG
Discriminant analysis
LORETA
BEAMFORMER
a b s t r a c t
Sixty-one dextral, unmedicated women with chronic fatigue syndrome (CFS) diagnosed according to the
Fukuda criteria (1994) and referred for investigation by rheumatologists and internists were studied with
quantitative EEG (43 channels) at rest with eyes open and during verbal and spatial cognitive activation. The
EEGs from the patients were compared with recordings from 80 dextral healthy female controls. Only those
subjects who could provide 20 1-s artefact-free segments of EEG were admitted into the study. The analysis
consisted of the identification of the spatial patterns in the EEGs that maximally differentiated the two
groups and the estimation of the cortical source distributions underlying these patterns. Spatial patterns
were analyzed in the alpha (8–13 Hz) and beta (14–20 Hz) bands and the source distributions were
estimated using the Borgiotti–Kaplan BEAMFORMER algorithm. The results indicate that the spatial patterns
identified were effective in separating the two groups, providing a minimum correct retrospective
classification rate of 72% in both frequency bands while the subjects were at rest to a maximum of 83% in the
alpha band during the verbal cognitive condition. Underlying cortical source distributions showed significant
differences between the two groups in both frequency bands and in all cognitive conditions. Lateralized
cortical differences were evident between the two groups in the both frequency bands during both the
verbal and spatial cognitive conditions. During these active cognitive conditions, the CFS group showed
significantly greater source-current activity than the controls in the left frontal–temporal–parietal regions of
the cortex.
Crown Copyright © 2009 Published by Elsevier Ireland Ltd. All rights reserved.
1. Introduction
Fukuda et al. (1994) have provided a modern operational
definition of chronic fatigue syndrome (CFS): 6 months or more of
increased fatigue, severe enough to reduce daily activity below 50% of
premorbid level with at least four of the following symptoms present
for at least 6 months or more: mild fever, painful lymph nodes,
myalgia, nonexudative pharyngitis, sleep disturbance (insomnia
or hypersomnia), migratory arthralgia, memory dysfunction, postexertional fatigue for more than 24 h. The above, of course, requires
that no primary medical or psychiatric illness be present that could
account for the symptomatology. Fibromyalgia (FM) differs from CFS
in that pain at particular trigger points is found, whereas fatigue
dominates the clinical picture in CFS, but the symptomatic overlap in
the two conditions is striking (Goldenberg, 1989). It has now been
well established that no single virus is etiological in CFS, but in 70% of
cases it is triggered by a viral “flu-like” illness. A preexisting immune
vulnerability is probable, given the overwhelming female suscepti-
⁎ Corresponding author. Department of Psychiatry, University of Alberta, Alberta
Hospital Edmonton, Box 307, Edmonton, Alberta, Canada T5J 2J7. Tel.: +1 780 472 5395;
fax: +1 780 472 5346.
E-mail addresses: [email protected],
[email protected] (P. Flor-Henry).
bility, women being more prone to autoimmune diseases than men,
for example, Hashimoto's thyroiditis, lupus, rheumatoid arthritis,
Sjogren's syndrome, antiphospholipid syndrome, Graves' disease/
hyperthyroiditis, scleroderma, myasthenia gravis, and multiple
sclerosis. That the immunologic system is involved is shown by
increase in activated T lymphocytes, increase in cytotoxic T cells,
increase in circulating cytokines and poor cellular function: low
natural killer cell cytotoxicity, poor response to mitogens in culture
and frequent immunoglobulin deficiencies: IgG1 and IgG3 (Landay
et al., 1991; Lloyd et al., 1988). In recent years it has become
increasingly recognized that the brain and the immune system
interact in a bidirectional chemical communication through neurotransmitters and cytokines modulating corticotropin releasing factor,
in turn regulating the hypothalamic–pituitary–adrenal axis (Black,
1995). One of the most important conveyors of immunological
information to the central nervous system (CNS) is the polypeptide
cytokine interleukin-1 (Ur et al., 1992). At the subjective level,
cognitive impairment is emphasized in patients with CFS, “mental
fog”, impaired concentration, poor memory and frequent word
finding difficulties. Numerous neuropsychological studies have
confirmed the presence of CNS dysfunction, essentially deficits in
information processing, memory impairment, and poor learning of
information (Marcel et al., 1996; Altay et al., 1990; Deluca and
Schmaling, 1995; Moss-Morris et al., 1996). Michiels and Cluydts
0925-4927/$ – see front matter. Crown Copyright © 2009 Published by Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.pscychresns.2009.10.007
156
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
(2001) and Tiersky et al. (1997) have reviewed these findings. Joyce
et al. (1996) further found that patients with CFS were also impaired
on verbal tests of unrelated word association learning and letter
fluency. Interestingly, Lange et al. (2005) found in a functional
magnetic resonance imaging (fMRI) study of verbal working memory
that although individuals with CFS could process auditory information
as accurately as controls, they utilized more extensive regions of the
network associated with the verbal working memory system and that
mental fatigue accounted largely for the increased blood-oxygenlevel-dependent (BOLD) signal change in the left superior parietal
region as well as increased activation bilaterally in the supplementary
and premotor regions. Numerous publications document a downregulation of the hypothalamic–pituitary–adrenal axis in chronic
fatigue (Parker et al., 2001; Demitrack et al., 1991; Demitrack, 1994;
Bearn et al., 1995; Poteliakhoff, 1981). Some 30–40% of CFS patients
have hypocortisolism. Abnormalities in serotonin neurotransmission
have been found in CFS: serotonin agonists lead to an increase in
serum prolactin in CFS that is not seen in depressed or healthy controls,
i.e. in CFS there is an up-regulation of serotoninergic neurotransmission, while in depression the opposite is found, hypercortisolism and
suppressed serotonin-mediated prolactin response (Demitrack and
Crofford, 1995). Excellent reviews of the general characteristics of CFS
have been provided by Wessely (1995), Afari and Buchwald (2003),
and Prins et al. (2006).
There are three diseases characterized by symptomatology that is
remarkably similar to that of chronic fatigue syndrome: Addison's
disease, Rickettsial diseases (Scientific American Medicine, 1987. 7:
XVII: 1–10), and glucocorticoid deficiency. For example, Addison's
disease is of sudden onset with an overrepresentation of middle-aged
women and presents with the following general symptoms: persistent
fatigue, debilitation after exercise, weakness, fever, enlarged lymph
nodes, myalgia, arthralgia, flu-like symptoms, sore throat, headaches,
and dizziness upon standing (Baschetti, 2000). Hypotension and
reduced aldosterone are an integral part of Addison's disease, and it is
probably not coincidental that postural hypotension (Bou-Holaigah
et al., 1995) and symptomatic relief with fludrocortisone (Florinef)
have been described in CFS. Remarkably, Scott et al. (1999)
administered the adrenocorticotropin (ACTH) stimulation test in 30
patients with CFS, in each case triggered by a viral influenza-type
illness, and found eight who had a blunted response to 1 μm
intravenously, i.e. failure to achieve a peak cortisol level greater than
600 nmol/l 30 min later. In all these subjects both adrenal glands
were 50% smaller than in controls on examination with computed
tomography (CT).
Some neurological illnesses may provide insight into the pathogenesis of CFS: measles encephalomyelitis is a neurological disorder of
abrupt onset that starts, on average, 5 days after the rash with
headaches, generalized seizures, confusion, and motor deficits. The
virus is not detected in the brain where the neuropathological
changes are similar to those seen in experimental allergic encephalomyelitis, suggesting an autoimmune pathogenesis (Johnson et al.,
1984). There was in the small locality of Akureyri in Iceland an
epidemic CNS illness in the winter of 1948–49 characterized by low
grade fever (not higher than 37.8 °C) with painful muscles, weakness,
nervousness and extreme fatigue. Seven years later, 75% had persistent
symptoms, 52% had muscle weakness and 65% had clear CNS signs
(Goodnick and Sandoval, 1993).
A number of investigations have reported reduced cerebral
circulation and hypometabolic changes in CFS, principally in the
frontal and temporal regions with single-photon emission computed
tomography (SPECT) and positron emission tomography (PET)
techniques. Sarkar and Seastrunk (1998) studied 150 patients, half
with CFS and half with idiopathic chronic fatigue, with SPECT and
magnetic resonance spectroscopy (MRS). With SPECT they found that
“the most reproducible area of perfusion abnormality is the left frontal
(or orbitofrontal) hypoperfusion in the cortical gray matter both in
early and delayed images on almost all patients”. MRS findings were
of decreased choline and increased lactate, again in the left frontal
lobe, correlating with SPECT hypoperfusion in that region. No such
changes were found in the right frontal lobe. In the temporal lobes
the metabolites were normal except for a mild lactate increase on
the left. Tomoda et al. (2000) reported on three children, ages 11, 12
and 13, with CFS that was triggered by a low grade flu-like illness.
The children were investigated with SPECT, which showed that the
left temporal and occipital blood flow was strikingly reduced in two
of the patients whereas in the third the blood flow in the left basal
ganglia and thalamus was increased. MRS revealed “a remarkable
elevation of the choline/creatine ratio”. Chaudhuri et al. (2003)
studied eight patients with CFS without psychiatric complications
with 1H-MRS and found a very significant increase in choline
compounds in the left basal ganglia. They did not, however, study
the right basal ganglia. Schwartz et al. (1993) find on SPECT imaging
of 45 patients with CFS that they had a similar number of regional
defects when compared with 14 patients with unipolar depression.
Tirelli et al. (1998), in a PET study (flurorodeoxyglucose, FDG) of 18
patients with CFS, reported significant hypometabolism in the right
medial frontal cortex and brainstem in these patients when
compared with healthy controls. Six psychiatric patients with
depression showed more severe hypometabolism bilaterally in the
medial and upper frontal regions, with normal brainstem. Machale
et al. (2000) compared 30 patients with CFS without depression, 12
psychiatric patients with major depression and 15 healthy controls,
measuring cerebral perfusion with SPECT. In patients with CFS and in
those with depression, there was increased circulation in the right
thalamus, pallidum and putamen, the CFS patients in addition
having increased perfusion in the left thalamus. Schmaling et al.
(2003) studied 15 patients with CFS and 15 healthy controls with
SPECT at rest and during the Paced Auditory Serial Addition test
(PASAT). Contrary to their hypothesis, there was greater brain
activation in the CFS subjects during the PASAT task, which was
more diffuse than in the controls, both groups performing equally
well on the test. Further, there was greater activation of the left
anterior cingulate in the CFS subjects. Siessmeier et al. (2003)
evaluated cerebral glucose metabolism with FDG-PET in 26 patients
with CFS (13 female; 13 male) and found that 12 of 26 patients were
similar to the controls, while 12 had bilateral hypometabolism of the
cingulate and mesial cortical regions. In addition, five of these had
decreased metabolism in the orbital frontal cortex. Anxiety and
depression, but not fatigue, were correlated with reduced glucose
metabolism. Interestingly, Ichise et al. (1992), given the overlap in
some of the symptoms found in depression and CFS, found no
differences in cerebral circulation in depressed and non-depressed
patients with CFS. Schwartz et al. (1994) compared MRI and SPECT
in 16 patients with CFS, finding that the magnetic resonance signals
were not statistically different from signals in 14 controls, although
the SPECT scanning revealed significantly more defects throughout
the cortex in the patients (present in 81% of patients and 21% of
controls). Abu-Judeh et al. (1998) came to similar conclusions in 18
patients with CFS, finding 13 abnormal SPECT results and 15 normal
FDG-PET results in the CFS patients. Abnormalities of cerebral
circulation may occur with normal glucose metabolism. The three
abnormal PET scans all had hypometabolism of the left parietal
region. Brooks et al. (2000), investigating 7 patients with CFS and 10
controls with 1H-MRS, found that the patients had a significant
reduction of N-acetylaspartate in the right hippocampus, even
although the volume of both hippocampi was similar to that in the
normal control subjects. Only the right hippocampus was studied.
Strikingly, the left hemisphere was implicated in many of the
imaging investigations reviewed above. Thus, we thought it would
be of interest to carry out a quantitative EEG analysis (current source
density) to see whether a similar lateralization might be found
electrophysiologically.
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
There have been few quantitative EEG investigations of CFS
published. Sherlin et al. (2007) reported on a LORETA study of
monozygotic twins discordant for CFS. They found that the CFS twins
had higher delta values in the left uncus and the left parahippocampal gyrus and higher theta values in the cingulate gyrus and the
right frontal gyrus. Their sample consisted of 17 pair of twins,
studied with 19 electrodes. Although their subjects were recorded in
the Eyes Open, Eyes Closed condition and during a mental arithmetic
task, they only reported the Eyes Closed data. Their twins were
unmedicated and they were administered the Diagnostic Interview
Schedule. Twenty-four percent of the CFS twins were depressed at
the time of study vs. none of the unaffected twins. Forty-seven
percent had an acute onset with a flu-like illness with an average
fatigue duration of 7.4 years and, characteristically, 88% were
female. The right hemisphere lateralization found here may reflect
the depressive aspect since there is evidence that the electrophysiological correlates of depression are of right fronto-temporal
excitation (Flor-Henry et al., 2004). Lewis et al. (2001), however,
in the study of 22 monozygotic twins, one of whom had CFS, found
that cerebral blood flow (SPECT) was the same in the affected and
unaffected twins. Guilleminault et al. (2006), in a sleep study of 14
patients with CFS, found in the CFS group a significant increase in
slow delta power relative to findings in normal controls. Siemionow
et al. (2004), in only eight CFS patients, observed at significant
increase in theta relative power during voluntary motor actions.
Loganovsky (2000) investigated the EEG characteristics of 100
randomly selected clean-up workers of the Chernobyl Accident, who
had been irradiated, and who presented with the symptomatology
of CFS as defined by the CDC criteria, which were met by 26. The
spectral analysis of the EEG showed lateralized (left-sided) increase
in theta power, decrease in alpha power and increase in beta power.
The somatosensory evoked potentials showed topographic abnormalities in the left temporo-parietal area. Scheffers et al. (1992)
found that the visual ERPs in a short-term memory task were similar
to those of normal controls in 13 cases, although mean reaction
times in the cases were slower. In the study of a single case, a 21 year
old woman, treated with neurofeedback and self-hypnosis, the
quantitative EEG showed increased left frontal theta activity
(Hammond, 2001). Tomoda et al. (2007) in the study of 414
children with CFS (ages 9–18) found prolonged P300 latency to
target stimuli in a subgroup increased P300 amplitude, and in a third
subgroup reduced P300 latency and amplitude.
Thus, these different approaches— neuropsychological, MRI, PET,
SPECT, immunological and endocrine — all show that in CFS central
nervous system functions are altered and that when the CNS
dysfunction is asymmetric, the left hemisphere is implicated. In an
attempt to further study these issues, we carried out a quantitative
EEG study of CFS. The alpha (8–13 Hz) and beta (14–20 Hz) bands in
the EEG were chosen for this study since these bands are thought to
reflect different aspects of cognitive activity: inhibitory and excitatory
brain functions, respectively (Pascual-Marqui et al., 1994).
157
tionnaires (Annett, 1970; Chapman and Chapman, 1987). All potential
control subjects were administered a pretest interview. Any who
responded in the affirmative to having a history of neurological
disease, birth complications, head trauma, psychiatric illness or
substance abuse were excluded from the study. In addition to the
pretest interview, the Quick Diagnostic Interview Schedule (Marcus
et al., 1989) was systematically administered to these subjects. Those
subjects with positive symptom categories were excluded from the
study. The Basic Personality Inventory (Jackson, 1996), which
compares the subject on 12 personality dimensions to the North
American female norms, and the Multidimensional Aptitude Battery
(MAB-II), which provides Verbal, Performance and Full Scale IQ, were
administered to all patients and controls (Jackson, 1998).
EEGs were recorded from all subjects during a passive condition,
eyes open (EO), and two active cognitive conditions, Word Finding
(WF) and Dot Localization (DL). At least 3 min of recording was
obtained from each subject. During the passive condition, the
subjects were told to relax, to think of nothing in particular and to
fixate on a point across the room. The tasks producing the cognitive
challenges were developed by (Miller et al., 1995) to be psychometrically matched, in terms of difficulty, item-scale correlations and
internal consistency, using a sample of 350 normal subjects. These
cognitive challenges were the same as those used by Davidson et al.
(1990) and Koles et al. (2001). The test questions were presented to
the subjects as slides projected from the rear of the recording
chamber. Once having formulated an answer to a question, the
subjects were asked to press a button to indicate that they were about
to verbalize a response. By using the recording of the button signal,
the EEGs were separated into cognitive and motor phases. Only the
cognitive phases were analyzed. Generally, the subjects required at
least 5 s to formulate their answers and neither the first nor the last
second of the EEG recorded during this period was selected for
analysis. Also, to ensure compliance, only those EEGs corresponding
to correct responses to the test questions were utilized. No rest
period was allowed the subjects between the questions.
The DL task contained 25 items with each item consisting of two
vertically arranged open rectangles, the top rectangle containing a
pair of dots and the bottom rectangle, slightly offset to either the right
or the left, containing an array of numbers. The offset of the bottom
rectangle differed randomly among items such that half of the items
were offset to either the right or the left side. Subjects were asked to
indicate which two numbers in the lower rectangle corresponded to
the location of the dots in the upper rectangle. Item difficulty is
determined by the size of the number arrays, which varied in five
levels from 8 to 50 numbers. In the WF task, subjects were asked to
correctly identify an object when presented with a written definition
of the object. For example, the correct answer for the item a very large
piece of floating ice is iceberg. The WF task consisted of 30 items that
differed in an ascending order of difficulty.
The research protocol outlined above was reviewed by the Ethics
Review Committee of the Alberta Hospital Edmonton and found to be
acceptable within the limitations of human experimentation.
2. Methodology
2.2. Recordings
2.1. Subjects
In order to further study the central nervous system characteristics
in chronic fatigue syndrome (CFS), quantitative EEG analysis was carried
out in a group of patients referred by rheumatologists, internists and
general practitioners. All satisfied Fukuda's criteria (1994), all were
dextral and were unmedicated at the time of study. In 67% of these
patients the onset was acute. Data were recorded from 61 women
suffering from CFS. The mean age of the group was 45.65 (S.D. ± 9.87).
Data were also recorded from a total of 80 female controls. The
mean age of this group was 25.57 (S.D. ± 8.37). These subjects were
completely right handed as determined by the handedness ques-
EEG recordings were obtained using a 43-lead stretchable electrode
cap containing tin electrodes. The 43 electrode sites represented
standard electrode positions described in the American Electroencephalography Society Guidelines (Sharbrough et al., 1990). The lead sites
were AF3, AF4, FC5, FC6, FC1, FC2, C3, C4, C1, C2, TP7, TP8, CP5, CP6, CP1, CP2,
P3, P4, P1, P2, PO3, PO4, AFz, Fz, FCz, Cz, CPz, Pz, POz, FP1, FP2, F7, F8, F3, F4, FT7,
FT8, T7, T8, P7, P8, O1, and O2. All leads were referenced to the left ear and
the impedance at each electrode was adjusted to be below 5 kΩ. Three
additional channels were recorded as follows: the electrocardiogram
(EKG), a bipolar recording over the external canthus to the supra-orbit
of the right eye to monitor vertical and horizontal eye movements
158
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
(EOG), and one bipolar electromyogram (EMG) channel recorded from
the temporomandibular region on the right and left sides. All channels
were amplified using a Grass Neurodata system and digitized with a 12bit RC Electronics analog-to-digital converter at a rate of 256 samples/s.
The band pass was set at 3–80 Hz with the high pass controlled by a
Grass Model 12A5 two-pole Butterworth filter with the 3-dB point set at
3 Hz. The anti-aliasing filter was an eight-pole Bessel function filter with
the 3-dB point set at 80 Hz.
2.3. Spectral analysis
The digitized EEG records were visually edited off-line in an
attempt to select from each subject 20 non-overlapping, 1-s data
segments corresponding to correct responses to the tasks. These were
selected randomly from all of the 1-s segments available from the
subject. Segments that contained voltage spikes, large eye movements, or muscle activity were not selected for analysis. Rejection of
segments was frequent and resulted in 58, 25 and 34 CFS patients
being able to contribute 20-segment EEGs during the EO, WF and DL
conditions, respectively. Of the control subjects, 79, 44 and 63 were
able to contribute 20-segment EEGs during these same conditions.
Only those patients and subjects who were able to contribute the 20
segments were admitted into the study.
The Fast Fourier Transform (FFT) was applied to each row of the
data matrix formed by the 46 EEG, EKG, EMG and EOG channels (as
rows) and 256 time samples (as columns) collected during each
second. Prior to the application of the FFT, each row of the data matrix
was preprocessed by applying a 50% Hamming taper (Brillinger 1981,
p.55). With 256 samples in each 1-s segment, the frequency
resolution of the FFT along the columns of each transformed data
matrix, V(f), was 1 Hz spanning the range of frequencies from 0 to
255 Hz. Since the alpha band (8–13 Hz) and low beta band (14–
20 Hz) were investigated in this study, only the columns 9 to 14 and
15 to 21 in V(f), corresponding to these frequencies, were selected.
The selected columns from the transformed data matrices were
characterized in terms of their cross-spectra. The cross-spectral
matrices Σ(α) and Σ(β) were formed as:
T
T
ΣðαÞ = VðαÞ V ðαÞ and ΣðβÞ = VðβÞ V ðβÞ
ð1Þ
where V(α) and Σ(β) are the alpha and beta columns from V(f) and T
is the complex conjugate transpose. An average 46 × 46 complexvalued cross-spectral matrix over the 20 segments available from each
subject was then calculated for each frequency band and for each of
the EO, WF and DL conditions.
The correlated effects of the EKG, EOG and EMG with the EEG were
removed using a matched filter (Brillinger, 1981 p.299) resulting in
43 × 43 cross-spectral matrices for each subject/frequency band/
condition. Each matrix was then normalized by dividing all of its
elements by its trace (the sum of its diagonal elements). This
procedure was implemented in order to remove the confounding
effects of skull thickness on EEG power (Wheeler et al., 1993).
2.4. Reduction of dimensionality
To reduce the effects of measurement noise and to minimize the
dependence of the EEG features captured in the cross-spectral
matrices on individual subjects, the dimensionality of the measurement space was reduced from 43 × 43 to 30 × 30 by factoring the sum
of the average cross-spectral matrices obtained from the CFS and
control groups in each frequency band/condition as:
T
ΣF + ΣC = E λ E
ð2Þ
where ΣF and ΣC are the average cross-spectral matrices for the CFS
and control groups, respectively, E is the matrix of eigenvectors and λ
is the diagonal matrix of corresponding eigenvalues arranged in
decreasing order. The reduction in dimensionality of the crossspectral matrices was obtained by using only the first 30 columns of E
to form:
T
T
ΣFR = ER ΣF ER and ΣCR = ER ΣC ER
ð3Þ
where ΣFR and ΣCR are the dimensionally reduced average crossspectral matrices of each group and ER contains the first 30 columns of
E. The number 30 was chosen arbitrarily but, in every case, the
retained components accounted for more than 99% of the combined
temporal variances in the cross-spectral matrices from the two
groups.
2.5. Separation of the groups
To assess the separation between the EEG features captured in the
reduced cross-spectral matrices obtained from the CFS and control
groups, the average cross-spectral matrix for each group were then
factored as (Koles et al., 2001):
T
T
ΣFR = C ΨF C and ΣCR = C ΨC C
ð4Þ
where C is the matrix of spatial patterns common to the two groups
and ΨF and ΨC are diagonal matrices containing the eigenvalues
corresponding to each of the spatial patterns. The spatial patterns in
the columns of C are determined from C = (P− 1)T where the matrix P
1
−1
contains the eigenvectors of either Σ−
FR ΣCR or ΣCR ΣFR both being the
same. The columns of C could be scaled such that ΨF + ΨC = I, the
identity matrix. With this scaling, it is evident that the spatial pattern
in C that accounts for the largest variance in the CFS group (as given by
the corresponding eigenvalue in ΨF) will account for the least
variance in the control group. Similarly, the spatial pattern in C that
accounts for the largest variance in the control group (as given by the
corresponding eigenvalue in ΨC) will account for the least variance in
the CFS group. Using these decompositions, a measure of the
Mahalanobis distance between the CFS and control groups was
calculated as:
2
−1
−1
D = traceðΣC ΣF Þ = traceðΨC ΨF Þ
ð5Þ
D2 in Eq. (5) can be interpreted as the sum of the squared distances,
measured in standard deviations, between the CFS and control groups
in the direction of each of the common spatial patterns. A D2 value of
30 would indicate that the CFS and control groups did not differ in
terms of the features captured in their reduced cross-spectral
matrices.
Fig. 1 shows the square roots of the elements of the diagonal
1
eigenvalue matrix Ψ−
C ΨF in Eq. (5) for the CFS and control groups
compared in the EO, WF and DL cognitive conditions and in both of the
alpha and beta frequency bands. The elements of ΨF were ordered
from largest to smallest (and hence ΨC from smallest to largest). All
six curves have the same general shape tending to flatten in the
middle and to pitch up and down at the left and right ends,
respectively. It is evident from the figure that the common spatial
patterns corresponding to the first and last eigenvalues account for
the greatest variance in one group compared with the other in all
cognitive conditions and in both frequency bands.
2.6. Classification of individuals
To assess the level to which the features contained in the crossspectral matrices obtained from the CFS and control groups would
separate the individual subjects in the groups, the classical complex
domain quadratic discriminant function was used to retrospectively
classify the subjects in each group. Under the assumption that the EEG
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
159
2.8. The BEAMFORMER transformation
Fig. 1. The Mahalanobis distance D between the CFS and q
control
groups contributed by
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
each of the elements of the diagonal eigenvalue matrix
are ordered from largest to smallest.
Ψ−1
C ΨF . The elements of ΨF
is a zero mean Gaussian process, this classifier is given by (Shumway,
1982):
−1
−1
−1
−1
T
QDF = traceððΣF −ΣC ÞΣi R Þ = traceðP ðΨF −ΨC Þ P Σi R Þ
ð6Þ
where ΣiR is an unclassified subject's reduced cross-spectral matrix.
Group centroids are obtained by setting ΣiR in Eq. (6) to ΣFR and ΣCR
respectively. This results in:
−1
−1
mF = traceðI−ΨC ΨF Þ and mC = traceðΨF ΨC −IÞ
ð7Þ
ΣiR is assigned to a group based on a decision line placed between the
centroids.
2.7. Estimation of cortical source-current density
The estimate of the cortical source-current density in each subject,
s2i , in each cognitive condition and frequency band was obtained
similarly to Koles et al. (2001):
2
T
T
T
T
si = diagðG ER C P ∑i R P C ER G Þ
ð8Þ
where G is the Borgiotti–Kaplan (BK) BEAMFORMER solution to the
EEG inverse problem (Sekihara et al., 2001). The diag operator in the
equation indicates that only the diagonal elements of the matrix
inside the brackets are used. G is a matrix with 43 rows and 51197
columns and s2i is a column vector with 51197 elements containing
the estimates of the magnitude of the source-current density at
locations in the cerebral cortex of the head model used for this study.
The matrix product CPT in Eq. (8) was included prior to the
BEAMFORMER transformation to eliminate the components in the
EEGs that were common to the CFS and control groups. To do this,
only those columns from C and P were retained that corresponded to
the largest and smallest values of ΨF (or the smallest and largest
values of ΨC). These columns were chosen because they account for
maximally different proportions of the temporal variances in the CFS
and control groups. The full matrix product CPT yields the identity
matrix; however, with only two selected columns, this is not the case
and, while the product matrix in this case has dimensions 30 × 30, its
rank is only 2. The reduced eigenvector matrix ER is used to transform
the 30-dimensional space back to 43 dimensions.
Transformation of the cross-spectral matrices, filtered with CPT and
ER, to source-current densities was implemented using the Borgiotti–
Kaplan Beamformer (Sekihara et al., 2001). The BK Beamformer differs
from the LORETA Transformation (LORETA-KEY®) in that it is not only
dependent on the lead-field matrix of the head model but also on the
cross-spectral matrix of the EEG. It has been shown (Russell and Koles,
2007) that the BK Beamformer is generally superior to LORETA in terms
of both the accuracy and the dispersion of source-current estimates.
The BK Beamformer was implemented in a realistic head model
obtained as a digitized MRI from the Brain Imaging Centre, Montreal
Neurological Institute (MNI). The model contains 4,079,350 cubic
elements and was segmented into five tissue compartments, scalp,
skull, CSF, gray matter, and other brain tissue. The elements contained in
each tissue compartment were assumed to be isotropic with conductivities assigned according to Haueisen et al. (1997). The solution space
consisted of 51,197 current dipoles distributed as a sheet in the gray
matter compartment of the head model. The lead-field matrix for the
head model was obtained using a variation of the preconditioned
conjugate-gradient method developed by Neilson et al. (2005). The MNI
head model was used, without any modification, to estimate the cortical
source-current density for of all of the subjects involved in this study.
To locate the positions of the electrodes on the head model, a
sphere that best approximated the scalp surface of the MNI model was
used to locate a center point for the sphere. Radial lines from the
center point were then projected toward the standard electrode
locations on the sphere and the electrodes placed on the model at the
points where the radial lines intersected the scalp surface. The same
electrode locations on the head model were used for all subjects.
Following the calculation of the estimates of the source-current
density vectors underlying the cross-spectral matrices for all of the
subjects, the elements of these vectors from the CFS and control groups
were statistically compared to identify brain regions that contained
statistically significant differences. The method used is based on
randomized t-tests of each element of the source-current density
vectors (Holmes et al., 1996). This method has the advantage of not
relying on underlying distributional assumptions while simultaneously
reducing the probability of obtaining false-positive results. The Holmes
method was used to determine the critical value of t, tc, such that if t N tc
the probability that there is no significant difference (two sided)
between the CFS and control groups is less than the chosen value of 0.05
(P b 0.05). The results of the statistical comparisons were summarized by
displaying the statistically significant t scores on a three- dimensional
rendering of the cerebral cortex of the head model.
3. Results
At the time of testing, the CFS patients were normal on all
personality dimensions except hypochondriasis, this reflecting the
somatic symptoms characteristic of CFS; notably, they did not indicate
depression (see Fig. 2). From the Quick Diagnostic Interview Schedule
it was found that over a lifetime 21 (or 34.4%) of the patients had
experienced depressive episodes, 7 (or 11.5%) social phobias, 10 (or
16.4%) agoraphobia, 9 (or 14.7%) generalized anxiety, 7 (or 11.5%)
panic attacks, 8 (or 13.1%) somatoform, 15 (or 24.6%) obsessive–
compulsive features, 2 (or 3.2%) bulimia, and 1 patient in the past had
experimented with stimulants, sedatives, cocaine, and heroin. Hickie
et al. (1990) evaluated the prevalence of psychiatric disorder in 48
patients with CFS. The premorbid prevalence of major depression was
12.5% and of total psychiatric disorder was 24.5%, which was not
higher than general community estimates. They further found that the
pattern of psychiatric symptoms in the CFS patients was similar to
what is seen in other medical disorders and different from what was
found in psychiatric patients.
160
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
Fig. 2. Basic Personality Inventory: women suffering from chronic fatigue syndrome versus women suffering from fibromyalgia.
In the female controls the mean verbal IQ was 115.5, performance
IQ 113.5, and full scale IQ 115.6. In the CFS patients, the verbal IQ was
111.2, performance IQ 111.9, and full scale IQ 112.1. There was no
statistically significant difference between the two groups. The women
with chronic fatigue are significantly older than the female controls
(Table 1). However, we expect that the normalization of the crossspectral matrix from each subject by its trace (total EEG power in the
band) will help to eliminate any dependence of EEG power on age.
Klimesch (1999), discussing alpha and theta oscillations that reflect
cognitive and memory functions in a review of the relevant literature,
concludes that with increasing age absolute power in the alpha and
theta band decreases whilst upper alpha power increases. Alpha
frequency increases from early childhood to adulthood and then
progressively decreases with age. Consequently, we analyzed the peak
alpha frequency of the patients and controls. Across all leads on
average the patients were 0.3 Hz slower than the controls — which was
non-significant (Hotelling T-squared value: 14.919; F = 1.14, d.f. =
(12 128), sig. = 0.33).
Fig. 2 shows the BPI profile of 63 women with chronic fatigue and 39
with fibromyalgia. (There are 2 extra women with CFS who were not
included in the study as they were tested after the analysis.) It is notable
that outside of the hypochondriasis scale (because of somatic
Table 1
Demographics of unmedicated, dextral patients suffering from chronic fatigue
syndrome versus dextral controls.
Age
Female controls
Female chronic Fatigue
Eyes open⁎
Mean
S.D.
Eyes closed⁎
Mean
S.D.
Dot localization⁎
Mean
S.D.
Word Finding
Mean
S.D.
N = 79
26.01
8.69
N = 80
25.57
8.37
N = 63
25.48
8.41
N = 44
25.82
8.02
N = 58
45.60
9.84
N = 61
45.65
9.87
N = 34
45.62
10.68
N = 25
45.44
9.59
⁎P b 0.01.
symptoms) both women with CFS and fibromyalgia are similar to
normal controls on 11 of the 12 dimensions. In particular neither the
fibromyalgia nor the CFS subjects were depressed at the time of testing.
The rate of correct classification of the CFS and control groups
based on only the two common spatial patterns that accounted for
maximally different proportions of the temporal variances between
the groups is shown in Table 2. To do this, only the largest and
smallest values of ΨF and the smallest and largest values of ΨC,
respectively, were inverted in Eqs. (4) and (5). The remaining values
in the inverse were set to zero. The decision line between the
centroids of the groups was placed so that the groups would be
classified at the same rate.
Table 2 shows the results of the classification. Surprisingly, based
on only two spatial patterns in a subject's EEG, the rate of correct
retrospective classification of the two groups is very high ranging
from a low of 72% of 130 subjects correctly classified in the EO
condition to a high of 83% of 64 subjects in the alpha band during WF.
Therefore, in the combined EO groups, 93 were correctly classified as
CFS or control. In the combined WF groups, 53 were correctly
classified as CFS or control. It would appear from the table that the WF
cognitive condition in the alpha band is slightly better at separating
the two groups than the beta band and that the active cognitive
conditions are better than the resting condition.
Figs. 3–8 show the statistical t (t N tc) maps of the cortical regions
were the source current density is significantly different (P b 0.05)
between the CFS and control groups in the alpha and beta bands and
in the three cognitive conditions. Shades of red indicate values of the t
Table 2
The percentage of the subjects in the CFS and control groups correctly classified using
only the two common spatial patterns that account for maximally different proportions
of the variance in the EEGs from the two groups in each of the alpha (8–13 Hz) and beta
(14–20 Hz) bands.
Condition
Alpha band
Beta band
Eyes open
Word finding
Dot localization
72%
83%
74%
71%
80%
74%
The decision line between the two group centroids was placed so that the subjects in
each of the two groups were correctly classified at the same rate.
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
161
Fig. 3. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the alpha (8–13 Hz) band during the eyes open (EO) cognitive condition.
Fig. 4. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the alpha (8–13 Hz) band during the word finding (WF) cognitive condition.
Fig. 5. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the alpha (8–13 Hz) band during the dot localization (DL) cognitive condition.
Fig. 6. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the beta (14–20 Hz) band during the eyes open (EO) cognitive condition.
162
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
Fig. 7. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the beta (14–20 Hz) band during the word finding (WF) cognitive condition.
Fig. 8. Statistical t maps (t N tc at P b 0.05) of the cortical regions where the source current density is significantly greater in the CFS group than in the control group (red hues) and
significantly greater in control group than in the CFS group (blue hues) in the beta (14–20 Hz) band during the dot localization (DL) cognitive condition.
statistic in regions where the source-current density is greater in the
CFS group than in the control group. Shades of blue indicate values of
the t statistic in regions where the source-current density is greater in
the control group than in the CFS group. A perusal of all the figures
shows a lateralization of the differences in source-current activity
between the groups in the WF and DL cognitive conditions. Elevated
source-current activity in the CFS group in the left frontal temporal
regions is particularly striking. This elevated activity is evident in the
alpha band during the WF cognitive condition and more so in the beta
band during the DL cognitive condition.
Table 2 indicates that the strongest separation between the CFS
and control groups in terms of the two common spatial patterns was
in the WF condition in both the alpha and beta bands. Examination of
the corresponding Figs. 4 and 6 reveals a strong left–right asymmetry
in the source-current density in both frequency bands with only the
alpha band showing significantly greater source-current activity in
the CFS group. However, a large region of elevated source-current
density in the left frontal–temporal–parietal regions of the CFS group
is very evident in the beta band (Fig. 8).
4. Discussion
The results related to the differences in the distribution of sourcecurrent densities between the CFS and control groups were obtained
using the Borgiotti–Kaplan (BM) BEAMFORMER. The algorithm was
modified, however, by first projecting each subject's cross-spectral
matrix on two spatial patterns (eigenvectors) the first of which
accounted for maximum variance in the EEGs of the CFS group and
minimum variance in the EEGs of the control group while the second
accounted for maximum variance in the EEGs of the control group and
minimum variance in the EEGs of the CFS group. If it is assumed that
each of these two spatial patterns corresponds to a single sourcecurrent distribution that best characterizes the difference between
one group and the other, it would seem then that the observed
differences between the groups can be interpreted in terms of the
activities of only two sources. Each source, however, consists of many
hundreds of individual dipole current sources that are temporally
correlated and spatially distributed over large contiguous areas of
cortex. Because large areas of cortex would appear to be involved in
CFS, single dipole localization methods would clearly not be suitable
for population studies of this kind. It has been shown (Russell and
Koles, 2007) that the BK Beamformer is generally superior to other
distributed source-localization algorithms in terms of both the
accuracy and the dispersion of estimated source-current distributions.
The essential findings in this investigation suggest that the features
in the EEG which best separate chronic fatigue syndrome from normal
brain function in females are the result of cortical source-current activity
that is generally suppressed in the alpha band and beta bands in the Eyes
Open condition in CFS. Further, during the Word Finding task there are
increased sources in the left frontal region in the alpha band and during
Dot Localization in the beta band. In both cognitive tasks, and in both
bands the sources of the right hemisphere are reduced. It is noteworthy
that the best classification, in both alpha and beta bands, is during the
Word Finding task. It has been generally accepted in the past that
decreased EEG power in the alpha band and increased power in the beta
frequency band correspond to increased functional activation in the
cerebral cortex. However, study of theta (4–8 Hz) and gamma (28–
64 Hz) oscillations show a significant increase in power during encoding
predicted subsequent recall (Sederberg et al., 2003). Hanslmayr et al.
(2009) reports that successful encoding in a semantic task is related to
power decrease in the alpha (8–12 Hz) as would be expected, but to
beta power decrease in the low beta (12–20 Hz) and increase in the
gamma band (55–70 Hz). Similarly, Sederberg et al. (2007), who
studied in epileptics with implanted intracranial electrodes during
verbal memory recall tasks report an increase in gamma band
oscillations (44–64 Hz) with a decrease in other bands. However, with
respect to the alpha frequency band the physiological meaning of alpha
oscillations is more complex. As Basar et al. (1997) comments: “Alpha as
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
a universal code or universal operator…the major physiological
meaning of 10 Hz oscillations which may be comparable to the putative
universal role of gamma responses in brain signaling.” The fact that
alpha does not necessarily reflect passive states or idling of the brain is
shown by the evoked 10 Hz oscillations that can be generated after
sensory stimulation and by event related alpha rhythms that can be
induced by cognitive processing. These findings are consistent with the
fMRI study of verbal working memory in CFS which shows increased
cerebral circulation in the left superior parietal region (and bilateral
activation in the premotor regions) in CFS during the processing of
complex verbal information (Lange et al., 2005). In view of these
complexities the precise interpretation of our findings is complex: in
line with the more generally accepted view that reduction of alpha
power corresponds to increased cortical activation, then in the Eyes
Open and during the Dot Localization condition the CFS group differs
from controls in having increased cortical activation bilaterally. During
Word Finding there is increased activation bilaterally, together with
reduced activation in the left frontal region. If one accepts these general
premises, in the beta band, the reduced bilateral sources in the Eyes
Open condition imply increased bilateral activation and in the Word
Finding condition the reduced sources in the right hemisphere similarly
imply increased activation of that hemisphere. In Dot Localization task
the right hemisphere sources are decreased and the left hemisphere
sources are increased over the whole of the left hemisphere, i.e.
activation of the right and extensive relative hypofunction of the left
hemisphere.
In a remarkable series of publications, Renoux and Biziere have
shown in right and left cortical lesioned rats, that in the mammalian
brain the T-cell immunological system is underlateralized in left brain
cortical control: with left hemisphere lesions there is a complete
collapse of T-cell mediated events, but this is under feedback inhibition
from the right hemisphere. Animals with right brain lesions have
normal T-cell immunological reactivity and are similar to healthy
animals; however,with left brain lesions, there is atrophy of the thymus,
whereas with right brain lesions there is hypertrophy of the thymus
(Renoux et al., 1983a,b; Biziere et al., 1985; Renoux and Biziere, 1986).
These findings were confirmed by Neveu (1988), who found that large
lesions of the left cerebral cortex depressed whereas lesions of the right
hemicortex enhanced T-cell immunity. Interestingly, Davidson et al.
(2003) studied 25 subjects before and after an 8-week training course in
meditation with quantitative EEG. These subjects and a control group
were then vaccinated with influenza vaccine. The meditators had a
significant increase in left brain activation (C3, C4) on an asymmetry
index in the 8–13 Hz band. Further, the meditators also had a
significantly greater rise in influenza antibody titers compared to the
compared with the controls that was significantly correlated to the
degree of left-sided anterior activation.
5. Conclusion
Thus, it would appear that dysregulation of the cortical hemispheric
controls of immunological functional systems (left hemisphere) and
hypothalamic–pituitary–adrenocortical axis is the underlying pathophysiology of CFS. The best discrimination from controls is during
cognitive activation of the left hemisphere during the Word Finding
task, over 80% correct (8–13 Hz). At the same time cognitive activation
of the right hemisphere in the spatial task, presumably through
transcallosal interaction, also induces increased sources in the left
hemisphere, (14–20 Hz). Quantitative EEG analysis would appear to be
a promising diagnostic indicator for this disorder.
References
Abu-Judeh, H.H., Levine, S., Kumar, M., El-Zeftawy, H., Naddaf, S., Lou, J.Q., AbdelDayem, H.M., 1998. Comparison of SPET brain perfusion and 18F-FDG brain
163
metabolism in patients with chronic fatigue syndrome. Nuclear Medicine
Communications 19, 1065–1071.
Afari, N., Buchwald, D., 2003. Chronic fatigue syndrome: a review. American Journal of
Psychiatry 160, 221–236.
Altay, H.T., Abbey, S.E., Toner, B.B., Salit, I.E., Brooker, H., Garfinkel, P.E., 1990. The
neuropsychological dimensions of postinfectious neuromyasthenia (chronic
fatigue syndrome): a preliminary report. International Journal of Psychiatry in
Medicine 20 (2), 141–149.
Annett, M., 1970. A classification of hand preference by association analysis. British
Journal of Psychology 61, 303–321.
Basar, E., Schurmann, M., Basar-Eroglu, C., Karakas, S., 1997. Alpha oscillations in brain
functioning: an integrative theory. International Journal of Psychophysiology 26, 5–29.
Baschetti, R., 2000. Chronic fatigue syndrome: a form of Addison's disease. Journal of
Internal Medicine 247, 737–739.
Bearn, J., Allain, T., Coskeran, P., Munro, N., Butler, J., McGregor, A., Wessely, S., 1995.
Neuroendocrine responses to d-fenfluramine and insulin-induced hypoglycemia in
chronic fatigue syndrome. Biological Psychiatry 37, 245–252.
Biziere, K., Guillaumin, J.M., Degenne, D., Bardos, P., Renoux, M., Renoux, G., 1985.
Lateralized neocortical modulation of the T-cell lineage. In: Guillemin, R., et al.
(Ed.), Neuroal Modulation of Immunity. Raven Press, New York, pp. 81–94.
Black, P.H., 1995. Psychoneuroimmunology: brain and immunity. Scientific American
Science and Medicine 16–25 November/December.
Bou-Holaigah, I., Rowe, P.C., Kan, J., Calkins, H., 1995. The relationship between neurally
mediated hypotension and the chronic fatigue syndrome. Journal of the American
Medical Association 274 (12), 961–967.
Brillinger, D.R., 1981. Time Series: Data Analysis and Theory. Holden-Day, San Francisco.
Brooks, J.C.W., Roberts, N., Whitehouse, G., Majeed, T., 2000. Proton magnetic resonance
spectroscopy and morphometry of the hippocampus in chronic fatigue syndrome.
British Journal of Radiology 73, 1206–1208.
Chapman, L.J., Chapman, J.P., 1987. The measurement of handedness. Brain and
Cognition 6, 175–183.
Chaudhuri, A., Condon, B.R., Gow, J.W., Brennan, D., Hadley, D.M., 2003. Proton magnetic
resonance spectroscopy of basal ganglia in chronic fatigue syndrome. NeuroReport
14, 225–228.
Davidson, R.J., Chapman, J.P., Chapman, L.J., Henriques, J.B., 1990. Asymmetrical brain
electrical activity discriminates between psychometrically-matched verbal and
spatial cognitive tasks. Psychophysiology 27 (5), 528–543.
Davidson, R.J., Kabat-Zinn, J., Schumacher, J., Rosenfranz, M., Muller, D., Santorelli, S.F.,
Urbanowski, F., Harrington, A., Bonus, K., Sheridan, J.F., 2003. Alterations in brain
and immune function produced by mindfulness meditation. Psychosomatic
Medicine 65, 564–570.
DeLuca, J., Schmaling, K.B., 1995. Neurocognitive testing in chronic fatigue syndrome.
Journal of Chronic Fatigue Syndrome 1 (3/4), 21–26.
Demitrack, M.A., 1994. Chronic fatigue syndrome: a disease of the hypothalamic–
pituitary–adrenal axis? Annals of Medicine 26, 1–5.
Demitrack, M.A., Crofford, L.J., 1995. Hypothalamic–pituitary–adrenal axis dysregulation in fibromyalgia and chronic fatigue syndrome: an overview and hypothesis.
Journal of Musculoskeletal Pain 3, 67–73.
Demitrack, M.A., Dale, J.K., Straus, S.E., Laue, L., Listwak, S.J., Kruesi, M.J.P., Chrousos, G.P.,
Gold, P.W., 1991. Evidence for impaired activation of the hypothalamic–pituitary–
adrenal axis in patients with chronic fatigue syndrome. Journal of Clinical
Endocrinology and Metabolism 73 (6), 1224–1234.
Flor-Henry, P., Lind, J.C., Koles, Z.J., 2004. A source-imaging (low-resolution electromagnetic tomography) study of the EEGs from unmedicated males with
depression. Psychiatry Research: Neuroimaging 130, 191–207.
Fukuda, K., Straus, S.E., Hickie, I., Sharpe, M.C., Dobbins, J.G., Komaroff, A., 1994.
International Chronic Fatigue Syndrome Study Group. The chronic fatigue
syndrome: a comprehensive approach to its definition and study. Annals of
Internal Medicine 121, 953–959.
Goldenberg, D.L., 1989. Fibromyalgia and its relation to chronic fatigue syndrome, viral
illness and immune abnormalities. Journal of Rheumatology 16, S91–S93.
Goodnick, P.J., Sandoval, R., 1993. Treatment of chronic fatigue syndrome and related
disorders: immunological approaches. In: Goodnick, P.J., Klimas, N.G. (Eds.),
Chronic fatigue and related immune deficiency syndromes. American Psychiatric
Press, Washington, D.C., pp. 109–129.
Guilleminault, C., Poyares, D., de Rosa, A., Kirisoglu, C., Almeida, T., Lopes, M.C., 2006.
Chronic fatigue, unrefreshing sleep and nocturnal polysomnography. Sleep
Medicine 7, 513–520.
Hammond, D.C., 2001. Treatment of chronic fatigue with neurofeedback and selfhypnosis. NeuroRehabilitation 16, 295–300.
Hanslmayr, S., Spitzer, B., Bauml, K.-H., 2009. Brain oscillations dissociate between
semantic and nonsemantic encoding of episodic memories. Cerebral Cortex 19,
1631–1640.
Haueisen, J., Ramon, C., Eiselt, M., Brauer, H., Nowak, H., 1997. Influence of tissue
resistivities on neuromagnetic fields and electric potentials studied with a finite
element model of the head. IEEE Transactions on Biomedical Engineering 44, 727–735.
Hickie, I., Lloyd, A., Wakefield, D., Parker, G., 1990. The psychiatric status of patients
with the chronic fatigue syndrome. British Journal of Psychiatry 156, 534–540.
Holmes, A.P., Blair, R.C., Watson, J.D., Ford, I., 1996. Nonparametric analysis of statistic
images from functional mapping experiments. Journal of Cerebral Blood Flow and
Metabolism 16, 7–22.
Ichise, M., Salit, I.E., Abbey, S.E., Chung, D.-G., Gray, B., Kirsh, J.C., Freedman, M., 1992.
Assessment of regional cerebral perfusion by 99Tcm-HMPAO SPECT in chronic
fatigue syndrome. Nuclear Medicine Communications 13, 767–772.
Jackson, D.N., 1996. Basic Personality Inventory Manual, 2nd ed. Sigma Assessments
Systems, Port Huron, MI.
164
P. Flor-Henry et al. / Psychiatry Research: Neuroimaging 181 (2010) 155–164
Jackson, D.N., 1998. Multidimensional Aptitude Battery II manual, 2nd ed. Sigma
Assessment Systems, Port Huron, MI.
Johnson, R.T., Griffin, D.E., Hirsch, R.L., Wolinsky, J.S., Roedenbeck, S., Lindo de Soriano, I.,
Vaisberg, A., 1984. Measles encephalomyelitis — clinical and immunologic studies.
New England Journal of Medicine 310 (3), 137–141.
Joyce, E., Blumenthal, S., Wessely, S., 1996. Memory, attention, and executive function
in chronic fatigue syndrome. Journal of Neurology, Neurosurgery and Psychiatry 60,
495–503.
Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory
performance: a review and analysis. Brain Research Reviews 29, 169–195.
Koles, Z.J., Flor-Henry, P., Lind, J.C., 2001. Low-resolution electrical tomography of the
brain during psychometrically matched verbal and spatial cognitive tasks. Human
Brain Mapping 12, 144–156.
Landay, A.L., Jessop, C., Lennette, E.T., Levy, J.A., 1991. Chronic fatigue syndrome: clinical
condition associated with immune activation. Lancet 338 (8769), 707–712.
Lange, G., Steffener, J., Cook, D.B., Bly, B.M., Christodoulou, C., Liu, W.-C., DeLuca, J.,
Natelson, B.H., 2005. Objective evidence of cognitive complaints in chronic fatigue
syndrome: a BOLD fMRI study of verbal working memory. NeuroImage 26,
513–524.
Lewis, D.H., Mayberg, H.S., Fischer, M.E., Goldberg, J., Ashton, S., Graham, M.M.,
Buchwald, D., 2001. Monozygotic twins discordant for chronic fatigue syndrome:
regional cerebral blood flow SPECT. Radiology 219, 766–773.
Lloyd, A.R., Wakefield, D., Boughton, C., Dwyer, J., 1988. What is myalgic encephalomyelitis? Lancet 1 (8597), 1286–1287.
Loganovsky, K.N., 2000. Vegetative-vascular dystonia and osteoalgetic syndrome or
chronic fatigue syndrome as a characteristic after-effect of radioecological disaster:
the Chernobyl accident experience. Journal of Chronic Fatigue Syndrome 7 (3),
3–16.
Machale, S.M., Lawrie, S.M., Cavanagh, J.T.O., Glabus, M.F., Murray, C.L., Goodwin, G.M.,
Ebmeier, K.P., 2000. Cerebral perfusion in chronic fatigue syndrome and
depression. British Journal of Psychiatry 176, 550–554.
Marcel, B., Komaroff, A.L., Fagioli, L.R., Kornish II, R.J., Albert, M.S., 1996. Cognitive
deficits in patients with chronic fatigue syndrome. Biological Psychiatry 40,
535–541.
Marcus, S., Robins, L.N., Bucholz, K., 1989. Quick Diagnostic Interview Schedule, III-R.
Steven Marcus, St. Louis.
Michiels, V., Cluydts, R., 2001. Neuropsychological functioning in chronic fatigue
syndrome. A review. Acta Psychiatrica Scandinavica 103, 84–93.
Miller, E.N., Fujioka, T.A.T., Chapman, L.J., Chapman, J.P., 1995. Psychometrically
matched tasks for assessment of hemispheric asymmetries of function. Brain and
Cognition 28, 1–13.
Moss-Morris, R., Petrie, K.J., Large, R.G., Kydd, R.R., 1996. Neuropsychological deficits in
chronic fatigue syndrome: artifact or reality? Journal of Neurology, Neurosurgery
and Psychiatry 60, 474–477.
Neilson, L.A., Kovalyov, M., Koles, Z.J., 2005. A computationally efficient method for
accurately solving the EEG forward problem in a finely discretized head model.
Clinical Neurophysiology 116, 2302–2314.
Neveu, P.J., 1988. Cerebral neocortex modulation of immune functions. Life Sciences 42,
1917–1923.
Parker, A.J.R., Wessely, S., Cleare, A.J., 2001. The neuroendocrinology of chronic fatigue
syndrome and fibromyalgia. Psychological Medicine 31, 1331–1345.
Pascual-Marqui, R.D., Michel, C.M., Lehmann, D., 1994. Low resolution electromagnetic
tomography: a new method for localizing electrical activity in the brain.
International Journal of Psychophysiology 18, 49–65.
Poteliakhoff, A., 1981. Adrenocortical activity and some clinical findings in acute and
chronic fatigue. Journal of Psychosomatic Research 25 (2), 91–95.
Prins, J.B., van der Meer, J.W.M., Bleijenberg, G., 2006. Chronic fatigue syndrome. Lancet
367, 346–355 Jan.28.
Renoux, G., Biziere, K., 1986. Brain neocortex lateralized control of immune recognition.
Integrative Psychiatry 4, 32–40.
Renoux, G., Biziere, K., Renoux, M., Guillaumin, J.M., 1983a. The production of T-cellinducing factors in mice is controlled by the brain neocortex. Scandinavian Journal
of Immunology 17, 45–50.
Renoux, G., Biziere, K., Renoux, M., Guillaumin, J.-M., Degenne, D., 1983b. A balanced
brain asymmetry modulates I cell-mediated events. Journal of Neuroimmunology
5, 227–238.
Russell, J.P., Koles, Z.J., 2007. A comparison of LORETA and the Borgiotti-Kaplan
Beamformer in simulated EEG source localization with a realistic head model.
International Journal of Bioelectromagnetism 9, 112–113.
Sarkar, S.M., Seastrunk, J.W., 1998. Develpment of radiology markers to assess clinically
diagnosed chronic fatigue syndrome. Nuclear brain SPECT and magnetic resonance
spectroscopy results. Rivista di Neuroradiologia 11 (Supp. 2), 60–65.
Scheffers, M.K., Johnson Jr., R., Grafman, J., Kale, J.K., Straus, S.E., 1992. Attention qnd
short-term memory in chronic fatigue syndrome patients: an event-related
potential analysis. Neurology 42, 1667–1675.
Schmaling, K.B., Lewis, D.H., Fiedelak, J.I., Mahurin, R., Buchwald, D.S., 2003. Singlephoton emission computerized tomography and neurocognitive function in
patients with chronic fatigue syndrome. Psychosomatic Medicine 65, 129–136.
Schwartz, R.B., Komaroff, A.L., Garada, B.M., Gleit, M., Doolittle, T.H., Bates, D.W., Vasile,
R.G., Holman, B.L., 1993. SPECT imaging of the brain: comparison of findings in
patients with chronic fatigue syndrome, AIDS dementia complex, and major
unipolar depression. American Journal of Roentgenology 162, 943–951.
Schwartz, R.B., Garada, B.M., Komaroff, A.L., Tice, H.M., Gleit, M., Jolesz, F.A., Holman, B.L.,
1994. Detection of intracranial abnormalities in patients with chronic fatigue
syndrome: comparison of MR imaging and SPECT. American Journal of Roentgenology 162, 935–941.
Scott, L.V., Teh, J., Reznek, R., Martin, A., Sohaib, A., Dinan, T.G., 1999. Small adrenal
glands in chronic fatigue syndrome: a preliminary computer tomography study.
Psychoneuroendocrinology 24, 759–768.
Sederberg, P.B., Kahana, M.J., Howard, M.W., Donner, E.J., Madsen, J.R., 2003. Theta and
gamma oscillations during encoding predict subsequent recall. Journal of
Neuroscience 23 (34), 10809–10814.
Sederberg, P.B., Schulze-Bonhage, A., Madsen, J.R., Bromfield, E.B., McCarthy, D.C., Brandt,
A., Tully, T.S., Kahana, M.J., 2007. Hippocampal and neocortical gamma oscillations
predict memory formation in humans. Cerebral Cortex 17 (5), 1190–1196.
Sekihara, K., Nagarajan, S.S., Poeppel, D., Marantz, A., Miyashita, Y., 2001. Reconstructing spatio-temporal activities of neural sources using an MEG vector beamformer
technique. IEEE Transactions on Biomedical Engineering 48, 760–771.
Sharbrough, F., Chatrian, G., Lesser, R.P., Luders, H., Nuwer, M., Picton, T.W., 1990.
Guidelines for standard electrode position nomenclature. American Electroencephalographic Society, Bloomfield, CT.
Sherlin, L., Budzynski, T., Kudzynski, H.K., Congedo, M., Fischer, M.E., Buchwald, D.,
2007. Low-resolution electromagnetic twins discordant for chronic fatigue
syndrome. NeuroImage 34, 1438–1442.
Shumway, R.H., 1982. Discriminant analysis for time series. In: Krishniah, P.R. (Ed.),
Handbook of statistics (Vol. II). North-Holland, Amsterdam, pp. 1–43.
Siemionow, V., Fang, Y., Calabrese, L., Sahgal, V., Yue, G.H., 2004. Altered central nervous
system signal during motor performance in chronic fatigue syndrome. Clinical
Neurophysiology 115 (10), 2372–2381.
Siessmeier, T., Nix, W.A., Hardt, J., Schreckenberger, M., Egle, U.T., Bartenstein, P., 2003.
Observer independent analysis of cerebral glucose metabolism in patients with
chronic fatigue syndrome. Journal of Neurology, Neurosurgery and Psychiatry 74,
922–928.
Tiersky, L.A., Johnson, S.K., Lange, G., Natelson, B.H., DeLuca, J., 1997. Neuropsychology
of chronic fatigue syndrome: a critical review. Journal of Clinical and Experimental
Neuropsychology 19 (4), 560–586.
Tirelli, U., Chierichetti, F., Tavio, M., Simonelli, C., Bianchin, G., Zanco, P., Ferlin, G., 1998.
Brain Positron tomography (PET) in chronic fatigue syndrome : preliminary data.
American Journal of Medicine 105 (3A), 54S–58S.
Tomoda, A., Miike, T., Yamada, E., Honda, H., Moroi, T., Ogawa, M., Ohtani, Y., Morishita,
S., 2000. Chronic fatigue syndrome in childhood. Brain & Development 22, 60–64.
Tomoda, A., Mizuno, K., Murayama, N., Joudoi, T., Igasaki, T., Miyazaki, M., Miike, T.,
2007. Event-related potentials in Japanese childhood chronic fatigue syndrome.
Journal of Pediatric Neurology 5, 199–208.
Ur, E., White, P.D., Grossman, A., 1992. Hypothesis: cytokines may be activated to cause
depressive illness and chronic fatigue syndrome. European Archives of Psychiatry
and Clinical Neuroscience 241, 317–322.
Wessely, S., 1995. The epidemiology of chronic fatigue syndrome. Epidemiologic
Reviews 17 (1), 139–151.
Wheeler, R.E., Davidson, R.J., Tomarken, A.J., 1993. Frontal brain asymmetry and
emotional reactivity: a biological substrate of affective style. Psychophysiology 30,
82–89.