RECORDING EEG DURING REPETITIVE TRANS-CRANIAL
MAGNETIC STIMULATION
Hubert de Bruin, Mark Archambeault, Trung Trinh and Gary Hasey
Dept. of Electrical and Computer Engineering, Dept of Psychiatry and Behavioural Neurosciences
McMaster University, Hamilton, Ontario, Canada
Keywords: rTMS, Stimulus artefact, Electrode heating, Artefact reduction, EEG, Trans-cranial magnetic stimulation,
Evoked potentials.
Abstract: This paper discusses several issues related to recording EEG during repetitive trans-cranial magnetic
stimulation (rTMS). The objective of recording EEG is to obtain magnetically evoked and event related
potentials. The issue of electrode heating is discussed and experimental results presented that show graphite
as well as fully notched or “C” silver, gold or silver-silver chloride are suitable for current rTMS protocols.
Standard silver or gold cup electrodes may cause excessive scalp heating. Removal or reduction of the
magnetically induced stimulus artefact is also discussed. A new system is presented that uses sample and
hold circuitry to block most of the artefact allowing the researcher to record ipsi- and contra-lateral evoked
potentials occurring within the first few milliseconds of the magnetic stimulus.
1 INTRODUCTION
Trans-cranial magnetic stimulation (TMS) has been
used during the past two decades to elicit responses
in the human brain. At first this modality was
suggested as a technique for studying upper motor
neuron health and function. This involved placing a
coil, either circular or “figure of eight” on the scalp
with its centre of strongest magnetic field over an
area of the motor cortex and exciting the underlying
cortical tissue with 300 – 400 μsec monophasic or
biphasic pulse. The response could be immediately
measured by recording the M-wave from the muscle,
usually the thenar, innervated by the upper motor
neurons under the coil. However, in recent years,
repetitive TMS has been proposed and used to treat
neuro-psychiatric disorders such as depression by
stimulating the medial frontal cortex (e.g. Fitzgerald
et al, 2003 and Hoffman et al, 2005). Since there are
no immediate recordable results such as the M-wave
for these sites, stimulus amplitudes have been
chosen by first determining the motor cortex
threshold and using a fraction, usually from 80 to
120% of this value. As well the stimulus site and
repetition frequency are chosen by convention rather
than patient responses. However, There is very
good evidence that the cortical responses are very
dependent on the stimulus site with some sites even
having no or very limited responses (Komssi et al,
2002). It has also been found that even the motor
threshold varies considerably intra-subject from
session to session (Wasserman, 2002) and there is
considerable evidence that frontal lobe thresholds are
higher than motor cortex thresholds (Kähkönen, 2005).
If one could record the immediate brain evoked
potentials (EP), or event related potentials (ERP),
stimulus amplitude, frequency and site could be
customized for each subject. This approach will also
allow us to gain valuable insight and knowledge
about the mechanisms of rTMS applied to the frontal
lobes. However, the very large magnetic fields
associated with 1 to 3 Tesla TMS pulses couple into
the patient electrodes, electrode cables and input
amplifiers resulting in very large voltage artefacts
that can saturate the input amplifiers for up to 500
ms. This paper discusses issues related to recording
EEG during rTMS and presents a new system for
recording magnetically evoked EPs and ERPs
2 ELECTRODE SELECTION
Not only does the very high magnetic field induce
currents in the cortex and deeper brain structures it
265
de Bruin H., Archambeault M., Trinh T. and Hasey G. (2009).
RECORDING EEG DURING REPETITIVE TRANS-CRANIAL MAGNETIC STIMULATION.
In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing, pages 265-272
DOI: 10.5220/0001534102650272
Copyright
c
SciTePress
also induces currents in the electrodes that are being
used to record the EEG. This induced current flow
will heat the electrodes. The temperature increase of
an electrode per stimulus is directly related to the
electrical conductivity of the electrode, the square of
the radius of the electrode and the square of the
stimulus strength (Roth, 1992). The conductivity of
materials tested is: Silver 62.9x10
6
S/m, Gold
41.0x10
6
S/m, Carbon 0.029x10
6
S/m (Serway,
2000), which suggests that silver electrodes will heat
the most per stimulus, while carbon electrodes will
heat the least per stimulus.
The safe temperature an electrode can reach
without causing cutaneous damage depends on the
exposure time. Figure 1 shows the time-surface
temperature threshold for first degree thermal injury.
Given that current rTMS treatment can include as
many as 2,400 pulses at a frequency of 0.25Hz up to
20Hz, electrode heating is a concern for causing
thermal skin damage. Looking at Figure 1, we see
that for a 30 minute study, 46°C is the hottest
temperature any scalp EEG electrode should be
allowed to heat to.
Figure 1: Time-temperature thresholds for burning of
human skin. Source: Moritz et al, 1947.
2.1 Electrode Testing
We decided to test a number of common and
modified EEG electrodes to determine which are
suitable for recording during rTMS. Temperature
was measured using a thermistor temperature probe
with 0.01°C accuracy (Digi-Sense LN5775, Cole-
Parmer, Illinois) calibrated to +/- 0.2°C and read to
0.1°C. The TMS machine used was a Magstim
Super Rapid. The TMS coil used was a Magstim
figure-of-eight air-cooled coil P/N 1640 (Inner
diameter 56mm, outer diameter 87mm, 9x2 turns,
16.4µH inductance, 0.93Tesla peak magnetic field).
Four commercially available EEG electrode types
were tested: (i) XLTEK reusable EEG/EP
electrodes Part no. #101339 (silver cup, 10mm
diameter, 2mm hole); (ii) Standard gold cup
electrodes (10mm diameter, 2mm hole); (iii)
Ag/AgCl surface electrodes model F-E5SCH-48
(10mm diameter, 2mm hole, Grass Technologies);
(iv) EL258RT reusable general purpose 8mm
diameter, no hole, radio-translucent carbon
electrodes with carbon leads (Biopac Systems Inc.).
Several gold and silver cup electrodes were pie-
notched and a gold plated silver cup electrode
(10mm diameter, 2mm hole, Nicolet) was fully
notched (C notched) to reduce induced current as
shown in Figure 2.
All testing used a sheet of plywood that was
marked with a 1cm grid pattern to help ensure
accurate coil placement. Electrodes were attached
using EEG paste (Ten20 conductive EEG paste).
The coil was placed so that the electrode was in the
area of maximum field induced heating as shown in
Figure 3. This position agreed with previous
findings (Roth et al 1992). The testing parameters
used in this study mirrored those used in standard
rTMS treatments: 10Hz stimulation for up to 8s
trains and 20Hz stimulation for up to 3s trains;
stimulus intensity was set at 85% of our machine’s
maximum to slightly exceed 110% of motor
threshold (MT) of an average subject at our own and
other rTMS laboratories (Thut et al, 2005).
Figure 4 shows the heating and cooling curves
for the 6 different electrodes tested with 3s 20Hz
trains at the maximum heating location (r = 30mm).
A second test was performed to simulate a full
treatment session where trains of 60 to 80 stimuli
are given at 1 minute intervals for a total of up to
3000 stimuli. Figure 5 gives the test results for a
gold cup electrode using only three trains of 60
stimuli at 20 Hz. As can be seen the electrode
temperature would soon rise above the maximum
allowable 46
o
C if more trains were given.
Carbon electrodes were also tested at 20Hz for
3s, 85% intensity, and 20Hz for 10s, 100% intensity
and showed 0.0°C and 0.8°C temperature rise
respectively. Repeating the 100% test with no
electrode resulted in 0.3°C temperature rise,
showing that the probe accounted for some of the
increase in temperature when 200 pulses were given.
Figure 2: Notched electrodes. From left to right: C
notched, pie notched, triple pie notched.
BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing
266
Figure 3: Stimulating coil showing how electrodes were
positioned underneath with respect to the r-axis, labeled in
cm.
Figure 4: Temperature effects for 6 different electrodes
from a single train of 3s at 20Hz at 85% intensity, r = 30
mm.
Figure 5: Temperature effects on gold cup electrodes from
3 trains of 3s at 20Hz at 85% intensity, r = 30 mm. Trains
were given at 0s, 60s, and 120s.
2.2 Electrode Guidelines
Unmodified silver and silver/silver chloride
electrodes appear unsuitable for standard rTMS-
EEG studies when high stimulus intensities are
necessary, due to the high conductivity of silver. If
used at all, pulse trains should not exceed 30 pulses
and electrodes should be allowed 290s to cool
between trains, given the current TMS and coil
parameters tested. Gold cup electrodes are suitable
for rTMS-EEG studies for a Magstim cooled coil if
stimulus intensity is kept below 85%, trains do not
exceed 80 pulses, and electrodes are allowed to cool
for 220s between stimulus trains. However,
notching does work, and when notched properly (a
full notch or C) electrode heating is reduced enough
to make silver and gold-plated silver electrodes
suitable for a standard rTMS-EEG study. The newly
available carbon electrodes should be suitable for
any rTMS-EEG patient study and their heating
would not be the limiting factor in selecting
stimulating parameters. However, they are about
three times the cost of Ag/AgCl electrodes.
3 ARTEFACT REDUCTION
Most modern commercial EEG systems are
protected from large voltage transients both by input
protection diodes and low pass filtering of the EEG
signal, typically below 70 Hz. Figure 6 shows the
average EEG recorded from a standard Ag cup
electrode at FP2 using a commercial EEG system
(XLTEK desktop EEG, Oakville, Ontario, Canada),
for six high amplitude TMS pulses delivered at the
F3 position. Although the amplifier hasn’t saturated
the artefact lasts at least 100 ms obscuring any EPs
and even some shorter latency ERPs. Several
researchers (Ives et al, 2006 and Fuggetta et al 2005)
have attempted to reduce this TMS stimulus artefact,
recorded using commercial EEG systems, by
designing low slew rate preamplifiers. Although
this approach is successful if only long latency ERPs
are considered, this low bandwidth is inadequate to
preserve the much higher frequency EPs. Further,
the sampling rate of commercial EEG systems (200
– 500 Hz) is too low to represent EPs.
Virtanen et al (1999) developed a multi-channel
EEG system with artefact blocking hardware to
record both EPs and ERPs following TMS.
However, this system cannot record EPs occurring
during the first 4 ms following stimulation and the
sampling rate of 1000 Hz is too low to fully capture
the shape of very short duration EPs. We have also
developed an artefact blocking system based on
sample and hold circuitry similar to Virtanen et al’s
approach. An earlier version (Archambeault and de
Bruin, 2007) was designed as a blocking
preamplifier for commercial EEG systems.
However, the sampling rates possible for these
RECORDING EEG DURING REPETITIVE TRANS-CRANIAL MAGNETIC STIMULATION
267
systems are inadequate for multichannel EP
recording.
3.1 A New System
We decided to implement a 16 channel EEG system
with the previous artefact blocking amplifiers using
the virtual instrument language Labview running on
a standard PC equipped with a National Instruments
DAQ interface. As shown in Figure 7, the DAQ
analog outputs are used to control the sample and
hold circuit for each channel and trigger the
magnetic stimulator. The system has selectable hold
time window duration and the stimulator trigger time
within this window. The user can also select the
channel sample rate, total signal period, stimulus
rate and the number of stimuli given. During rTMS
the channel recordings are continually displayed for
analysis and verification, and synchronously
averaged for background EEG and instrumentation
noise reduction. Following completion of the
stimulus train the averge signal for each channel is
stored in an EXCEL format file for further signal
processing and analysis using programs such as
Matlab. This gives us a very flexible clinically
friendly system with aggregate sampling rates up to
200 KHz (at least 10 KHz per channel). This system
can easily be upgraded to 32 channels with different
DAQ hardware.
Subject 1
-40
-30
-20
-10
0
10
20
0.00.51.01.52.0
Time (s)
EP Voltage (µV)
Figure 6: Averaged ERP, created from six ERPs recorded
with XLTEK EEG machine. TMS over left hemisphere
frontal cortex. EEG recorded from right hemisphere
frontal cortex.
Ch0+
Gnd
Sample
andhold
Unity
gain
buffer
HPF
.16Hz
GainStage
(Non
inverting
opamp,
LPF
1500Hz
LabVIEW
DAQboard
Analog
Output
Computer
Magneticstimulator
Ch0‐
GainStage
(Instrumentation
Amplifier,10x
gain)
Figure 7: Artefact blocking EEG machine design using the sample-and-hold (blocking) approach (one channel shown).
BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing
268
Figure 8: EEG responses for medial left frontal lobe
stimulation at F3 at 66% maximum amplitude.
Figure 9: EMG responses for left thenar muscle for two
recording electrodes, Ag and AgCl for wrist stimulation at
90% maximum amplitude.
3.2 System Tests
The system was tested to determine whether the
blocking circuit could manage large magnetic
artefacts and which hold window durations were
suitable. A 24 year old subject was instrumented
with Grass 10 mm Ag/AgCl electrodes at F3 and F4,
the right mastoid (reference) and the neck (ground).
60 magnetic pulses at a rate of 2 Hz were given near
F3 in the medial frontal lobe at 66% of the
maximum amplitude of a Magstim Super Rapid
stimulator. The hold time duration was set to 2.6
msec with the stimulus trigger command given at 0.3
msec after the start of the sample and hold
command. The signals were collected for 100 msec
at 5 KHz sample rate. Figure 8 shows the average
EEG signals for both locations.
In this figure the first rectangular pulse is the
offset voltage applied by the sample and hold circuit
and can be viewed as the total hold window.
Unfortunately both Labview and the Magstim
control program are Windows based and there is
some uncertainty that the magnetic stimulus is
given at the precise time (0.3 msec after the start of
hold). The maximum artefact block in this case is
2.3 msec but it could be less. The negative
excursion is due to the residual stimulus artifact,
which decays exponentially toward the baseline.
The F3 signal shows the evoked muscle M-wave
resulting from magnetic stimulation of the
temporalis muscle under the coil. This usually
occurs within 1 msec of the magnetic motor point
stimulation. F4, as expected shows no muscle
response and only the decaying artefact, which is
almost the same size as the F3 artefact. Both signals
were heavily contaminated by 60 Hz and other
environmental noise and, although not synchronized
to the stimuli, were still not entirely removed by
averaging. The laboratory contained a number of
high power instruments with large transformers
resulting in very large 60 Hz ambient noise.
Because of the residual noise in the signal we cannot
be sure the small μvolt excursions were brain evoked
potentials.
The duration and amplitude of the artefact are
determined by stimulus amplitude and shape, as well
as the impedances of the electrodes, electrode wires
and input amplifier. If these impedances were
purely resistive, the artefact would last no longer
than the stimulating pulse (400μsec). We wanted to
address the issue whether polarisable or non-
polarisable electrodes would result in less artefact.
The hypothesis was that a non-polarisable electrode
such as Ag-AgCl would store less artefact energy
because of their low capacitance. Ag or Au
commercial polarisable electrodes, on the other hand
would store more energy resulting in a larger
residual artefact when the sample and hold circuit
reconnected the electrodes to the amplifier.
The left thenar muscle of a 24 year old male was
instrumented with two electrodes, Ag-AgCl and Ag,
in close proximity. A reference Ag-AgCl electrode
distal to the second joint of the thumb and a ground
Ag-AgCl electrode on the dorsum of the hand. The
Magstim figure-of-eight coil was placed at the wrist,
5 cm equidistance from the two thenar electrodes.
The sample and hold circuitry was set to block the
RECORDING EEG DURING REPETITIVE TRANS-CRANIAL MAGNETIC STIMULATION
269
artefact a maximum of 1.7 msec after the stimulus
was given . A single 90% maximum amplitude
stimulus pulse was given by the Magstim and 100
msec of signal recorded from both electrodes at 5
KHz sample rate. Figure 9 shows the first 50 msec
of the unprocessed signals.
As for Figure 8, the hold window of 2.0 msec
appears first, followed by a rise to approximately
4.25 mV (amplifier saturation) due to the residual
stimulus artefact. The resulting signal excursion is
due to the evoked muscle M-wave and the decaying
stimulus artefact. The signals after 15 msec are due
to 60 Hz and other environmental noise. The M-
waves and decaying stimulus artefacts are very
similar with the Ag-AgCl signal decaying slightly
faster. This could also be due to the slightly
different M-waves recorded by the two electrodes.
Lower levels of stimulation that resulted in much
smaller M-waves showed the same similarities. At
this point it must be concluded that the type of
electrode has little effect on the residual stimulus
artefact and our early hypothesis that non-
polarisable electrodes would have lower
magnetically induced artefacts was wrong.
3.3 Multi-Channel Tests
The multi-channel system was tested for 16 channels
of evoked potentials recorded from a male subject
instrumented using the standard 10-20 electrode
configuration. The human tests were approved by
the Research Ethics Board of St. Joseph’s Health
Care, Hamilton, Ontario, Canada. Figure 10 shows
the averaged responses for 80 stimuli given at 8 Hz
at 69% of the Magstim maximum amplitude (110%
of the motor threshold) with the coil placed over
Brodmann area 46. The signals were sampled at 5
KHz and bandpass filtered from 15 Hz to 2.5 KHz.
At the scale shown all that can be seen are the very
large amplitude muscle responses or M-waves from
the underlying temporalis and occipitofrontalis
muscles. The null response for F3 is a result of
amplifier saturation since the coil was placed over
F3 and the amplifier hold time was only 2 msec.
Figure 11 shows the same response starting at 15
msec with increased resolution. Fp1 shows EPs at
18 msec, while other EPs can also be seen at 47 and
85 msec in all channels. These synchronous EPs
may be a result of using linked reference electrodes
over the mastoid bone.
Figure 10: EEG averaged responses for 80 stimuli at 8 Hz,
69% max, Brodmann area 46. One unit = 1 mV, channels
spaced by 5 mV.
Figure 11: EEG averaged response of Fig. 10 One unit =
0.25 µV, channels spaced by 25 µV.
Figure 12: EEG averaged responses for 80 stimuli at 8 Hz,
69 % max, Brodmann area 9. One unit = 0.25 µV with
channels spaced by 25 µV.
BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing
270
Figure 12 shows the results for the same stimulus
train but with the coil placed over Brodmann area 9.
Two features can be noted: (i) the F3 channel is not
saturated since the coil was not immediately over the
F3 electrode position and the pattern of EPs is
different for that shown in Figure 11.
4 CONCLUSIONS
Our research has shown that standard commercial
Au and Ag or Ag-AgCl EEG electrodes cannot be
used for general rTMS applications, due to excessive
skin heating. However, these electrodes can be used
if fully or “C” notched. Further the choice among
these electrodes does not seem to affect the
amplitude or duration of the stimulus artefact.
Although our heating results are in general
agreement with the conclusions of previous
researchers, the lack of dependency of the stimulus
artefact amplitude on electrode material does not
(Virtanen et al, 1999). Their results show that the
amplitude depends mostly on electrode size and that
Ag-AgCl electrodes had very low artefacts
compared to Ag, although this could be a result of
the very small Ag-AgCl pellet size. Further, their
fully notched Ag standard electrodes had much
lower artefact than the intact ones. The principle
contributors to the stimulus artefact are not well
understood, and electrode, wire and input amplifier
capacitances all play a part. Even in a multichannel
recording situation, where the magnetic field
orientation and amplitude is very different for each
electrode, the residual stimulus artefacts can be very
similar as shown in Figure 8. Further research will
be conducted to investigate the determining factors
for magnetically induced artefacts and how common
these are for all stimulating and recording
conditions.
The new EEG system works very well, and
depending on the stimulus strength, the amplifier
can be reconnected to the recording electrodes with
delays from 1 to 4 msec, allowing us to record EPs
as well as ERPs. The initial voltage offset
introduced by sample and hold circuitry, shown as
the square pulse in Figures 8 and 9, can be ignored
and does not affect the signal when the block is
terminated. Tests with the stimulating coil at some
distance from the electrodes resulting in very low
short duration artefacts have shown that the
amplitude returns to baseline within μsec. The test
results shown in this paper were for worst case
scenario experiments with large stimulus amplitudes
and the coil either directly over the recording
electrodes or in close proximity. However, the
system must be made more immune to
environmental noise by better shielding and cabling.
Future work will include postprocessing to
estimate and remove the exponentially decaying
residual artefact and periodic environmental noise.
Although synchronous averaging can remove
asynchronous environmental noise if enough stimuli
are given, the technique is inefficient. The number
of stimuli presented to the brain should be
determined by clinical efficacy rather than noise
reduction.
The multi-channel results show that even when
stimulus artefact is removed new signal processing
techniques will have to be developed to model and
remove the muscle M-waves.
ACKNOWLEDGEMENTS
Support for this research from the Natural Sciences
and Engineerung Research Council of Canada is
gratefully acknowledged.
REFERENCES
Archambeault, M., de Bruin, H. 2007, Development of an
electroencephalography data acquisition system for
clinical research into transcranial magnetic stimultion
evoked potentials, Proc. IEEE Cdn. Conf. ECE,
Vancouver, pp 202 – 205.
Fitzgerald P.B., Brown, T.L., Marston, N.A., Daskalakis,
Z.J., De Castella, A., Kulkarni, J. 2003, Transcranial
magnetic stimulation in the treatment of depression: a
double-blind, placebo-controlled trial. Arch Gen
Psychiatry 60(10): 1002-1008.
Fuggetta, G., Fiaschi, A., Manganotti, P. 2005, A direct
demonstration of cortical oscillatory activities induced
by varying single-pulse transcranial magnetic
stimulation intensity over the left primary motor area:
a combined EEG and TMS study. Neuroimage 27(4):
896-908
Hoffman, R.E., Gueorguieva, R., Hawkins, K.A.,
Varanko, M., Boutros, N.N., Wu, Y.T., Carroll, K.,
Krystal, J.H. 2005, Temporoparietal transcranial
magnetic stimulation for auditory hallucinations:
safety, efficacy and moderators in a fifty patient
sample. Biol Psychiatry 58(2): 97-104.
Ives, J.R., Rotenburg, A., Poma, R., Thut, G., Pascual-
Leone, A. 2006, Electroencephalographic recording
during transcranial magnetic stimulation in humans
and animals. Clin Neurophysiol 117:1870-1875.
RECORDING EEG DURING REPETITIVE TRANS-CRANIAL MAGNETIC STIMULATION
271
Kähkönen, S., Komssi, S., Wilenius, J., Ilmoniemi, R.J.
2005, Prefrontal TMS produces smaller EEG
responses than motor-cortex TMS: implications for
rTMS treatment in depression. Psychopharmacology
181:16-20
Komssi, S., Aronen, H.J., Huttunen, J., Kesäniemi, M.,
Soinne, L., Nikouline, V.V., Ollikainen, M., Roine,
R.O., Karhu, J., Savolainen, S., Ilmoniemi, R.J. 2002,
Ipsi- and contralateral EEG reactions to transcranial
magnetic stimulation. Clin Neurophysiol 113:175-184.
Roth, B.J., Pascual-Leone, A., Cohen, L.G., Hallett, M.,
1992, The heating of metal electrodes during rapid-
rate magnetic stimulation: a possible safety hazard.
Electroencephalography and Clinical Neurophysiol
85: 116-123.
Serway, R. A., Beichner, R. J. 2000, Physics For
Scientists and Engineers. Orlando: Saunders College
Publishing,.
Thut, G., Ives, J.R., Kampmann, F., Pastor, M.A., Pascual-
Leone, A. 2005, A new device and protocol for
combining TMS and online recordings of EEG and
evoked potentials. Journal of Neuroscience Methods
141: 207-217.Virtanen, J., Ruohonen, J., Näätänen, R.,
Ilmoniemi, R.J. 1999, Instrumentation for the
measurement of electric brain responses to transcranial
magnetic stimulation. Med Biol Eng Comput 37:322-
6.
Wassermann, E.M. 2002, Variation in the response to
transcranial magnetic brain stimulation in the general
population. Clin Neurophysiol 113:1165-1171
BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing
272