Transcript Slayt 1

COST B27 ENOC Joint WGs Meeting
Swansea UK, 16-18 September 2006
BOLD Changes During Driven
Electrical Oscillations
in Human Brain
Ahmet Ademoğlu
Bogazici University,
Institute of Biomedical Engineering,
Istanbul, Turkey
BOLD Changes During Driven Electrical Oscillations in Human
Brain
Zübeyir Bayraktaroğlu1, Uzay E. Emir2, Cengizhan Öztürk2, Ahmet Ademoğlu2, Tamer Demiralp1
1 Istanbul University, Istanbul Faculty of Medicine, Department of Physiology,
2 Bogazici University, Institute of Biomedical Engineering, Istanbul, Turkey
Introduction:
While EEG represents the spatial summation of the synchronous electrical
activity of neurons (real functional signal of neural activity), with a high
temporal resolution, it is insufficient for localisation of the brain structures
that generate these signals. On the other hand, fMRI BOLD (Blood
Oxygene Level Dependent) response reflects perfusion and oxygen level
changes in the brain tissue with a high spatial resolution, but its’ temporal
resolution is not enough to follow neural dynamics.
The correct modeling of the neurovascular coupling is still a very important
gap for revealing the relationship between the BOLD response and
functional neural activity patterns. Tonic neural discharges can be
expected to generate increases in BOLD response, whereas it is not yet
systematically investigated how the metabolic activity changes during
oscillatory activities in the EEG.
A good way of producing synchronization patterns in the EEG that are
stationary within the time-constant of the BOLD response is to evoke
steady-state evoked potentials. When the brain is stimulated with stimuli
at a high repetition rate, steady-state evoked potentials are obtained in the
EEG at the stimulation frequency and its’ harmonics (Regan, 1989). When
steady-state evoked potentials to visual stimuli within the range of 1-100
Hz were investigated, amplitude increases have been found at certain
stimulation frequencies (10, 20, 40 and 80 Hz) close to the peaks in the
EEG spectrum (Herrmann, 2001). Although the functional significance of
this frequency selectivity is not yet fully understood, it is considered that
they reflect the dynamic properties of the neural networks responsible
from different stages of sensory processing.
In the preliminary stage of the present study, we investigated EEG and
BOLD responses to steady-state visual stimulation with a checkerboard
reversal pattern created by a computer. To overcome the refreshing rate
limitations of the graphics card and the data projector and to increase the
frequency range and resolution of the visual stimuli, we developed a LED
(light emitting diode) based device with fiberoptic transmission system for
the MRI room. This stimulus presentation system enabled us to apply
visual stimuli within the 1-100 Hz frequency range with 1 Hz steps.
Materials and Methods:
In this study, the BOLD responses during diffuse light flickering at rates
between 1-100 Hz have been systematically studied and the change of
the BOLD response has been analyzed in relation to the stimulus
presentation frequency.
fMRI – BOLD responses were recorded from 8 healthy subject (6 male, 2
female), aged 28.25 ± 5.06 years.
Figure 1. BOLD percentage change with increasing stimulation
frequency.
1 Hz
2 Hz
3 Hz
4 Hz
5 Hz
6 Hz
7 Hz
8 Hz
9 Hz
10 Hz
11 Hz
12 Hz
13 Hz
14 Hz
16 Hz
18 Hz
20 Hz
22 Hz
24 Hz
28 Hz
32 Hz
36 Hz
40 Hz
44 Hz
50 Hz
60 Hz
70 Hz
80 Hz
90 Hz
100 Hz
fMRI recordings:
BOLD measurements were conducted with a 1.5 Tesla Siemens Syngo
MRI System using a single shot T2* weighted gradient echo planar
imaging sequence. A 3D MPRAGE sequence was used for high resolution
anatomic scan. The stimuli were presented at 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
11, 12, 13, 14, 16, 18, 20, 22, 24, 28, 32, 36, 40, 44, 50, 60, 70, 80, 90
and 100 Hz. 3 minutes of fMRI recording has been taken for each
flickering frequency.
Recording Basal BOLD response (30 sec)
Recording BOLD response during stimulation (30 sec)
Experimental design blocks
Analysis:
fMRI data were processed by AFNI software package. Images were
registered in order to remove motion artifact. For each dataset, activated
regions were determined by correlating one regressor general linear
model.
Results:
In the preliminary study, it has been observed that the BOLD response
showed two peaks at 5 and 10 Hz. The decreasing BOLD amplitude with
increasing frequencies built another peak around 18 Hz. These results are
consistent with the current fMRI literature [Parkes, 2004].
In the second phase of the study, diffuse light was used for stimulation.
This allowed to see effects of stimulation frequency in a better resolution.
BOLD signal did not show a linear increase or saturation with increasing
frequencies, but displayed certain peaks and deeps at certain frequencies
(Figure 1).
Figure 2. The anatomic localization and amount of BOLD response change with
increasing stimulation frequency.
Changes in the localization and extent of the activation also varied
between frequencies. The BOLD response peaked around 5, 10, 16, 36
and 60 Hz stimulation frequencies. At these stimulus frequencies a
tendency for wider activation regions was observed.
Discussion:
The fact that the BOLD response did not show a linear increase or
saturation with increasing frequencies, but displayed peaks around certain
frequencies that roughly correspond to the frequencies at wich steadystate potentials show higher amplitudes, points to the presence of a
relationship between the metabolic activity and the electrical oscillations in
the EEG.
Additionally, the wider distribution of BOLD activations at these
frequencies suggests that these preferred stimulus rates are more
effective in activating the neighbouring secondary areals around the
primary visual cortex.
References:
1. Herrmann CS (2001). Human EEG responses to 1–100 Hz flicker: resonance phenomena in visual
cortex and their potential correlation to cognitive phenomena. Exp Brain Res 137, 346–353
2. Logothetis NK, Pfeuffer J (2004). On the nature of the BOLD fMRI contrast mechanism. Magnetic
Resonance Imaging 22, 1517–1531
3. Parkes LM, Fries P, Kerskens CM, Norris DG (2004). Reduced BOLD response to periodic visual
stimulation. NeuroImage 21, 236– 243
4. Regan D (1989). Human brain electrophysiology: evoked potentials and evoked magnetic fields in
science and medicine. Elsevier, New York
POLYMORPHISMS OF DRD4 AND DAT1 MODULATE
HUMAN GAMMA BAND RESPONSES
Tolgay Ergenoglu4, Christoph S. Herrmann2, M. Emin Erdal3, Yasemin H. Keskin1,
Mehmet Ergen1, Hüseyin Beydagi4, Tamer Demiralp1
(1) Istanbul University, Istanbul Faculty of Medicine, Department of Physiology, Turkey
(2) Otto-von-Guericke University Magdeburg, Department of Biological Psychology, Germany
(3) Mersin University, Medical Faculty, Department of Medical Biology and Genetics, Turkey
(4) Mersin University, Medical Faculty, Department of Physiology, Turkey
INTRODUCTION
Electrophysiological recordings in different species indicate that
gamma oscillations (30-70 Hz) in the brain are associated with
a variety of fundamental perceptual and cognitive processes
[1]. Possible generation mechanisms have been proposed for
gamma oscillations [2, 3]. However, less is known about the
neurochemical basis of the modulation of evoked gamma
responses during cognitive processes, although attention and
working memory tasks modulate them [4].
Significant changes in the gamma responses have been
observed in schizophrenia and Attention Deficit Hyperactivity
Disorder (ADHD), which lead to significant failure of attentional
modulation and to working memory deficits [5]. Both disorders
also have significant associations with three genetic
polymorphisms concerning the dopamine system, which is
critical for cognitive functions. Polymorphisms of the dopamine
receptor D4 (DRD4) gene, dopamine transporter (DAT1) gene
and catechol-Omethyltransferase (COMT) gene showed
significant associations with schizophrenia and ADHD [6,7].
Therefore, we aimed to investigate whether direct relations
exist between the DRD4, DAT1 and COMT polymorphisms and
the amount of gamma oscillations of normal subjects evoked by
an auditory selective attention paradigm.
MATERIALS AND METHODS
Subjects: Fifty right-handed, healthy, male aged 21.5 ± 1.64
years
Electrophysiological Recordings: EEG and ERPs were
derived from 16 electrodes placed according to the 10-20
system (Oz, O1, O2, Pz, P3, P4, Cz, C3, C4, T3, T4, Fz, F3, F4,
Fp1, and Fp2). EOG was recorded for artifact elimination.
Cognitive Paradigm: A classical auditory oddball paradigm
was employed.
Analysis of ERPs and Oscillations: After rejection of artifacts,
peak amplitudes and latencies of the P50, N100, and P300
waves of the averaged ERPs were measured. For the analysis
of event-related oscillations, the data were transformed to the
time-frequency plane using a wavelet transform.
Wavelet Transform: To compute a wavelet transform, the
original signal was convolved with a complex Morlet wavelet.
Evoked acitivity phase-locked to stimuli was calculated by
applying WT on average responses of each subject.
Genotyping:
Venous blood sample collection into ethylene diamine tetra
acetic acid (EDTA)
DNA extracted from peripheral blood leukocytes by salting
out procedure
The polymerase chain reaction (PCR) based genotyping of
the polymorphisms
The PCR products resolved on a 2.5 % agarose gel
containing 0.5µg/ml ethidium bromide
The gel visualized under UV light using Vilber Lourmat
system
Polymorphisms and genotypes:
DRD4 exon III
: 2/2, 2/4, 3/4, 4/4, 4/6, 2/7, 3/7, 4/7, 7/7
DAT1 VNTR
: 8/10, 9/9, 9/10, 10/10
COMT Val 108/158 Met
: H/H, H/L, L/L
Statistics: The amplitude and latency differences of ERPs and
amplitude differences of evoked gamma responses between
groups with different genotypes were tested by a repeated
measures ANOVA design with the genotype as the between
subjects factor (genotype: 2 levels - 7 repeat vs others for
DRD4, homozygous 10/10 vs. others for DAT1 and
homozygous H/H vs others for COMT), and stimulus (2 levels:
standard vs. target), anteroposterior topography (3 levels:
frontal, central, parietal) and lateral topography (3 levels: left,
midline, right) as the within-subject factors.
List of detected genotypes and number of subjects with each
genotype is demonstrated in table.1.
ERP Results:P50, N100, or P300 amplitudes or latencies were
not significantly different between the 2 groups of genotypes for
any polymorphism.
Evoked gamma band responses: The analysis of the DRD4
polymorphisms revealed a significant increase of gamma
activity for the 7-repeat allele (genotype 2) both for target and
standard stimuli (genotype: F(1,46)=10.66, p<0.01). For the
DAT polymorphism an interaction of the factors genotype and
stimulus indicated that only targets are influenced by this
polymorphism (genotype x stimulus type interaction:
F(1,46)=4.33, p<0.05). In the group with genotype 2, the target
gamma response was significantly higher than in the subjects
with genotype 1 (genotype: F(1,46)=4.62, p<0.05). COMT
genotypes revealed no effect on gamma activity. (Figures.1 & 2)
Figure 1. Time-frequency representations of the ERPs to
auditory stimuli (average of target and standards) in frontal
midline location (Fz) for genotypes 1 and 2.
DRD4
DAT1
2/2, 2/4, 3/4, 4/4, 4/6
(n=38)
8/10, 9/9, 9/10
(n=25)
H/H
(n=15)
Genotype 2
2/7, 3/7, 4/7, 7/7
(n=10)
10/10
(n=23)
H/L, L/L
(n=33)
Table.1: Distribution of genotypes.
The homozygous 10-repeat allele (10/10) of the DAT1
polymorphism introduced a significant amplitude increase
specifically in evoked gamma responses to targets,
whereas no significant change was observed in gamma
responses to standards. The inefficient variant of DAT1
that yielded the enhanced gamma response to targets
probably also resulted in enhanced dopamine levels in
extracellular space. Because it has been proposed that
task-related activity in neurons of prefrontal cortex (PFC)
during working memory is modulated by dopamine
mainly via the D1 receptor [8], it seems plausible to
assume that this DAT1 effect was mediated through the
action of increased extracellular dopamine on the D1
receptor.
The absence of any differences between the evoked
gamma responses of the subjects with the high and lowactivity variants of the COMT gene seems to be
contradictory to the results obtained with DAT
polymorphism. However, the facts that the uptake by the
DAT is the most effective mechanism for the termination
of the synaptic action of dopamine in the brain and that
the role of COMT remains minimal under normal
conditions [9] could explain this difference between the
DAT and COMT results.
In conclusion, our results suggest that the action of
dopamine via the D4 receptor inhibits the evoked gamma
response nonselectively to all stimuli. However,
increased levels of extracellular dopamine, due to an
inefficient DAT, selectively enhance target gamma
responses and probably reflect the D1-mediated
dopaminergic contribution to a prefrontal target detection
mechanism.
REFERENCES
Figure 2. Time courses of evoked gamma activity in response
to auditory stimuli in Fz for genotypes 1 and 2.
RESULTS
Because 2 of the 50 subjects had a high number of trials with
artifacts, they were excluded from further analyses. Each of the
DRD4, DAT1 and COMT polymorphisms were divided into two
subgroups according to the associations of the genotypes with
cognitive disorders. Genotype 2 always shows some
association with cognitive disorders.
Genotype 1
DISCUSSION
In our study, the 7-repeat isoform of DRD4 polymorphism
yielded a significant increase in the auditory evoked
gamma responses to both target and standard stimuli.
This finding is in line with the gamma results [5] and
DRD4 results in ADHD [6], which showed a significant
association between ADHD and the 7-repeat allele of the
DRD4 polymorphism. The D4 receptor can affect
potassium channels as well as GABAergic chloride
channels [3] thus modulating the excitability of neurons.
Generally, dopamine is believed to inhibit activity of
pyramidal cells if effective via the D4 receptor. Increased
gamma activity in subjects with the 7-repeat isoform of
DRD4 polymorphism might be the result of less inhibition
via the D4 receptor.
COMT
Figure 3. Topographical distribution of the evoked gamma
activity in the time interval from 40 to 60 ms for genotypes 1 and
2. Responses are maximal over frontal electrodes
(anteroposterior: F(2,94)=6.33, p<0.01). The increase of
gamma oscillations for the 7-repeat allele (genotype 2) of the
DRD4 polymorphism and 10/10 genotype (genotype 2) of the
DAT polymorphism are also maximal over frontal electrodes.
1. Basar E, Schürmann M, Basar-Eroglu C, Demiralp T (2001) Selectively
distributed gamma band system of the brain. Int. J. Psychophysiol., 39, 129135.
2. Gray CM, König P, Engel AK, Singer W (1989) Oscillatory response in the cat
visual cortex exhibit intercolumnar synchronization which reflects global
stimulus properties. Nature 338, 334-337.
3. Traub RD, Jefferys JG, & Whittington MA (1999) Fast oscillations in cortical
circuits. MIT press.
4. Herrmann CS, Mecklinger A (2001) Gamma activity in human EEG is related
to highspeed memory comparisons during object selective attention. Vis.
Cogn., 8, 593-608.
5. Yordanova J, Banaschewski T, Kolev V, Woerner W & Rothenberger A (2001)
Abnormal early stages of task stimulus processing in children with
attentiondeficit hyperactivity disorder--evidence from event-related gamma
oscillations. Clin.Neurophysiol., 112, 1096-1108.
6. Faraone SV, Doyle AE, Mick E, Biederman J (2001) Meta-analysis of the
association between the 7-repeat allele of the dopamine D(4) receptor gene
and attention deficit hyperactivity disorder. Am. J. Psychiatry, 158, 1052-1057.
7. Herken H, Erdal ME (2001) Catechol-O-methyltransferase gene polymorphism
in schizophrenia: evidence for association between symptomatology and
prognosis. Psychiatr. Genet. 11, 105-109.
8. Seamans JK, Durstewitz D, Christie BR, Stevens CF, Sejnowski TJ (2001)
Dopamine D1/D5 receptor modulation of excitatory synaptic inputs to layer V
prefrontal cortex neurons. Proc Natl Acad Sci USA. 98, 301-306.
9. Huotari M, Santha M, Lucas LR, Karayiorgou M, Gogos JA, Mannisto PT
(2002) Effect of dopamine uptake inhibition on brain catecholamine levels and
locomotion in catechol-O-methyltransferase-disrupted mice. J. Pharm. Exp.
Therap., 303, 1309-1316.
Subtopographic EEG Source Localization After
Spatio-Temporal Wavelet Decomposition
Duru A. D*1., Eryilmaz H1., Bayram A1., Ademoglu A.1, Demiralp T. 2
1 Biomedical
2 Department
MOTIVATION
Localization of the cognitive activity in the brain is one of the major problems in
neuroscience. Current techniques for neuro-imaging are based on fMRI, PET and EEG
recordings. The highest temporal resolution is achieved by EEG, which is crucial for
temporal localization of activities. But spatial resolution of scalp topography for EEG is
low.
To overcome the spatial resolution limitation of scalp topography, several currentdensity estimation techniques were developed.
The goal is to find the location of the three-dimensional (3D) intracerebral activities by
solving an inverse problem.
EEG generally consists of several electrical sources some of which are temporally as
well as spatially overlapping. For this reason the scalp topologies constituted by these
multiple sources makes the inverse problem more complicated.
Engineering Institute, Bogazici University, Istanbul, Turkey
of Physiology, Istanbul Medical School, Istanbul University, Turkey
SIMULATION
Dipole Source Configuration of Simulated EEG
Temporal
Characteristics
Superficial Delta (3 Hz)
Spatial Characteristics
Source 1
Source 2
Deeper Delta (3 Hz)
Low Spatial Frequency Map
Source 3
Superficial Delta (3 Hz)
High Spatial Frequency Map
Source 4
Deeper Delta (3 Hz)
Low Spatial Frequency Map
Source 5
Superficial Alpha (14 Hz)
High Spatial Frequency Map
Source 6
Deeper Alpha (14 Hz)
Low Spatial Frequency Map
Source 7
Superficial Alpha (14 Hz)
High Spatial Frequency Map
Source 8
Deeper Alpha (14 Hz)
Low Spatial Frequency Map
Source Position
The aim of the spatio-temporal decomposition of EEG scalp maps by wavelet transform
is twofold;
High Spatial Frequency Map
BEM
Figure 4) 64 channel Simulated Total EEG activity
for the 8 Sources in defined in Table 1. Sampling
Rate is 256 Hz and the duration is 1 s.
Topography
Table 1) Source configuration for simulated EEG
data.
Source Localization (MUSIC)
i) to isolate temporal frequency components of EEG into bands like delta, theta, alpha
...
Topography
ii) to isolate the scalp maps of these individual components into spatial frequency maps
which are determined by the depth and extension of individual sources prior to their
source localization.
Figure 5) Topography of Total EEG at 180 th time
sample.
5 octave Temporal Decomposition
METHODS
Figure 6) Music Spectrum of Total EEG data given
in Fig. 5.
Source Localization
Delta right deeper Alpha right superficial
Realistic Head Model
5 octave Spatial Decomposition and Localization for delta and alpha bands
data
Delta right deeper
Alpha right deeper
Delta left superficial Alpha left superficial
Delta left deeper
Alpha left deeper
Figure 1) MRI images. (177x240x256, voxel size 1mm x 1mm x
1mm)
The head model that we used in this study is developed using the average T1 weighted
human brain MRI data provided by Montreal Neurology Institute (MNI). Statistical
Parameter Mapping software 99 release (SPM99) which is developed by Wellcome
Institute is used for 3-D segmentation of the brain, skull and scalp. After segmentation,
the surfaces are triangulated in order to generate the realistic head model that we need
to solve the forward problem.
REAL DATA
Real EEG data is obtained from Istanbul University, Istanbul Medical School. Twenty-four healthy right-handed volunteers (13 males and 11 females) were recruited as subjects with a mean
age of 25.8 ± 5.6 and a mean education of 17.8 ± 3.3 years. The CPT paradigm consisted of 400 stimuli, 10 distractors (B, C, D, E, F, G, H, J, K, L), 1 primer “A”, 1 target “Z” appearing with
the following probabilities: 20% primers, 10% Go stimuli (any “Z” after an “A”), 10% NoGo stimuli (any distractor letter after an “A”) and 60% distractors. EEG was amplified with a band pass of
0.1–70 Hz from 30 scalp electrodes, Oz, O1, O2, Pz, P3, P4, P7, P8, Cz, C3, C4, T7, T8, Fz, F3, F4, FCz, FC3, FC4, CPz, CP3, CP4, FT7, FT8, F7, F8, TP7, TP8, FP1, FP2, and sampled at
200 Hz. After building the ERP epochs of 1500 ms duration between −500 and 1000 ms, trials with EEG or EOG amplitudes exceeding ±90 μV were rejected automatically as artifact. ERPs
were averaged for the Go and NoGo CPT paradigms
Temporal Decompositon (Delta Coefficient 3 (350-525ms))
Figure 2) Tesellated a) brain, b) skull and c) scalp surfaces tesselated with 2000, 1000, 1016 triangles, respectively.
Temporal Decompositon (Theta Coefficient4 (525-700 ms))
Figure 7) 30 channel averaged ERP activity for Go and
NoGo CPT respectively. Sampling Rate is 200 Hz and
the duration is 1.5 s.
CPT GO
CPT NOGO
EEG Electrode Registration for Simulation and Processing
5 octave Spatial Decomposition and Localization for D3
CPT GO
CPT NOGO
5 octave Spatial Decomposition and Localization T4
Figure 3) a) Position of 64 electrodes, b)The surface of scalp registered with electrodes.
Blue colored points shows the electrode positions.
64 channel EEG electrode locations are registered to the scalp surface by spline
interpolation using the T1 weighted MR data, the inion-nasion and pre-auricular
coordinates, and the 5-10 Electrode Placement which is similar to the International 10-20
Electrode Placement System. The surface of the scalp is densely represented by 16188
triangles for registration and topographic mapping (Fig 3).
Forward Problem
Forward problem of EEG, which computes the electrical potentials on the scalp surface
given the source positions and strenghts, is solved using the
Boundary Element Method (BEM) with the Center of Gravity (COG) approximation on a
realistic head model given in Fig 2.
Inverse Problem
The inverse problem, which estimates the source positions and their strength from
multichannel EEG data, is solved using the Multiple Signal Classification (MUSIC)
scanning algorithm. MUSIC is based on subdividing the brain tissue into a 3-D grid and
computing the spatial power spectrum with an eigenbased approach for each voxel
element. In this study, gray matter is scanned as a solution space for MUSIC algorithm
with a voxel grid of 8mm x 8 mm x 8 mm.
CONCLUSIONS
The T4 coefficient that represents the theta response between 525 and 700 ms shows a left lateralized activation in temporally decomposed data. After spatial decomposition we obtain a
clear activation on the left motor cortex that probably corresponds to the motor activity related with the button press with the right hand in addition to a cerebellar activation. The same timefrequency region in the NoGo condition shows two separate activations after spatial decomposition: One in the posterior parietal area and another activation in prefrontal cortex, which does
not appear at all in the source localization of the temporally decomposed data. The orbito-frontal activation might correspond to the response inhibition in the NoGo condition of the CPT
paradigm.
The D3 coefficient corresponding to the delta response between 350 and 525 ms mostly resembling the topgraphy of the Go-P3 wave is represented with a single source in the temporaly
decomposed data of the Go condition, whereas after spatial decomposition multiple generators appear in the mesial surface of the posterior parietal cortex and in left frontal area as expected
for the Go-P3. The same delta coefficient in the NoGo condition seems to be generated by a parietal and two bilateral temporal generators when the raw data is used for source localization.
After spatial decomposition two generators in the left frontal region appear in addition to a sharper dissociation of the parietal and bilateral temporal sources.
Temporal wavelet analysis of EEG at a given spatial location yields temporally stationary components at temporal frequency bands like delta, theta, alpha. Spatial wavelet analysis of EEG at
a given temporal location yields spatially stationary scalp maps at spatial frequency bands. The characteristics of these maps are determined by the depth and extension of individual EEG
sources. A spatiotemporal preprocessing of the EEG simplifies the complexity of the scalp map by separating it into several submaps each of which is produced by an individual EEG
source. This is a very convenient preprocessing prior to source localization for the isolations of different maps corresponding to different dipole sources. This way, even the temporally
correlated EEG sources can be localized after spatio-temporal decomposition of EEG.