Acquisition of human EEG data during linear self-motion on a Stewart platform

Abstract

The present study investigated the feasibility of acquiring electroencephalography (EEG) data during self-motion in human subjects. Subjects performed a visual oddball task - designed to evoke a P3 event-related potential - while being passively moved in the fore-aft direction on a Stewart platform. The results of this study indicate that reliable EEG data can be obtained during self-motion on a Stewart platform: this finding is important for the ecological validity of further research into human motion.

Full text

Page 1

Acquisition of Human EEG Data during
Linear Self-Motion on a Stewart Platform

Hugh Nolan, Robert Whelan , Richard B. Reilly
Trinity Centre for Bioengineering
Trinity Institute for Neuroscience
Trinity College Dublin, Ireland
[nolanhu, robert.whelan, richard.reilly]@tcd.ie
Heinrich H. Bülthoff, John S. Butler
Max Planck Institute for Biological Cybernetics
Tübingen, Germany
[heinrich.buelthoff,
john.butler]@tuebingen.mpg.de

Abstract— The present study investigated the
feasibility of acquiring
(EEG) data during self-motion in human subjects.
Subjects performed a visual oddball task – designed
to evoke a P3 event-related potential – while being
passively moved in the fore-aft direction on a Stewart
platform. The results of this study indicate that
reliable EEG data can be obtained during self-motion
on a Stewart platform: this finding is important for
the ecological validity of further research into human
motion.
electroencephalography
Keywords—
Stewart platform; vestibular; multisensory; P3
electroencephalography; Self-motion;
I.
INTRODUCTION
A. Background
Multisensory integration (MSI) is an essential
aspect of human function that is, as yet, poorly
understood. Despite the fundamental role that
sensory integration plays in performance and
perception, how and when information from
separate sensory modalities comes together in the
human neocortex is an unsolved problem. The
human sense of movement and balance integrates
vestibular information with visual [ 1 ] and
somatosensory information [2].
Understanding visual-vestibular sensory integration
is key in understanding human balance. This is
important because processing of vestibular signals
deteriorates with age, which can cause postural
instability and falls. Epidemiological studies show
that falls are a very common problem in the older
population with 30% of community dwelling
elderly people >65 years of age falling each year
and 12% of these falling at least twice [3][4][5].
Falls are the leading cause of injury-related
hospitalization in over 65’s and hip fractures are
the leading cause of mortality due to injury in
people aged over 75 years [6]. Therefore, there is a
need for a simple non-invasive method of
measuring the vestibular system in vivo.
B. Imaging methodologies for MSI
There are many complexities involved in acquiring
neural data from subjects who are moving. For
example, neither magnetic resonance imaging
(MRI) nor positron emitted tomography (PET) are
suitable methods as the subject must remain in a
fixed position during the procedure. Therefore,
research on the neural correlates of linear self-
motion in humans has typically used visual vection,
which is self-motion induced by large-field visual
motion stimulation during which the stationary
subject perceives the moving visual surroundings
as being stable and themselves as moving [7][8].
Electroencephalography (EEG) would appear to be
a suitable candidate for recording neural activity
during motion because modern EEG acquisition
equipment is lightweight
Furthermore, a Stewart motion platform [ 9 ]
provides an appropriate method of producing linear
self-motion in a laboratory environment. Up until
now, study of true self-motion in humans has been
limited to rotational motion [10][11].
There are a number of potential drawbacks to using
a Stewart motion platform in conjunction with EEG
recording equipment. For example, noise could be
introduced into the EEG signal from the motion of
the actuators, the electrical noise of the platform
power source, or from muscular activity of the
subject as they compensate for the acceleration at
the start and finish of the motion.
and portable.
C. Experimental paradigm
In an attempt to address the potential issues
described above, we first investigated the
feasibility of recording EEG during motion by
running a standard EEG task - an oddball paradigm
- with four different motion conditions to address
each of these potential issues.
An oddball paradigm is one in which occasional
target stimuli have to be detected in a train of
frequent irrelevant standard stimuli. It is known to
evoke robust cognitive event-related potentials
(ERPs; time-locked EEGs), which provide an
objective and non-invasive tool for measuring the
time-course of neural
paradigms evoke a P3 response [12] which is a
positive ERP component, with a parietocentral
scalp distribution, which typically peaks from 250
to 600 ms following an infrequent stimulus. The P3
has been investigated in a wide range of neurologic
disorders, primarily because it can be reliably
elicited with relatively simple paradigms.
processing. Oddball
SaD1.9
Proceedings of the 4th International
IEEE EMBS Conference on Neural Engineering
Antalya, Turkey, April 29 - May 2, 2009
978-1-4244-2073-5/09/$25.00 ©2009 IEEE585
Authorized licensed use limited to: TRINITY COLLEGE LIBRARY DUBLIN. Downloaded on August 28, 2009 at 04:28 from IEEE Xplore. Restrictions apply.

Page 2

The four different levels of motion were termed
static, idle, slow and fast. The static condition
served as a control as there was no interference
from the experimental setup and participant
motion. In the idle condition the motors raised the
platform to its neutral position – this is a
centralized position that allows a full range of
movement in all directions. The motors require
power to maintain this position; thus allows
investigation into the influence of electrical noise
of the motors on the EEG signal. In the slow and
fast conditions participants were sinusoidally
moved forward and backwards at a frequency of
0.5Hz. These conditions are of most interest as
there may be interference from the setup and use of
the electrical motors, but also from the motion of
the participant.
II.
METHODS
A. Participants
Six male participants with normal or corrected-to-
normal vision completed the experiment for
payment. All participants except for the authors,
RW, JB and HN, were naïve to the purpose of the
experiment. The average age was 25 years (range
18-31). Participants gave their informed consent
before taking part in the experiment, which was
performed in accordance with the ethical standards
laid down in the 1964 Declaration of Helsinki.

B. Apparatus
The experiment was conducted in the Motion Lab
at the Max Planck Institute for Biological
Cybernetics, Tübingen, Germany. The main piece
of equipment used was the Maxcue 600 platform
manufactured by Motion-Base PLC [13]. This is a
6-legged platform controlled by hydraulics. It has 6
degrees of freedom, which are rotation and
translation about 3 axes. The visuals were
displayed on a projection screen, with a field of
view of 86°×65° and a resolution of 1400×1050
with a refresh rate of 60 frames per second.
Participants responded using a four-button response
box.
EEG data were recorded using the BrainAmp
MRplus™ [14] electrode system from 8 scalp
electrodes along the scalp midline and one EOG
channel, digitized at 5000 Hz. The system has a
hardware highpass filter at 0.16Hz and a hardware
lowpass filter at 1000 Hz which were applied
during data acquisition. EEGLAB was used for all
EEG analyses [15]. The EEG data from the Pz
channel were downsampled to 512 Hz, bandpass
filtered between 2–45 Hz, epoched from -1000 to
2000 ms after stimulus presentation, and baseline
corrected from 200
presentation. Data were manually scanned for
obvious artifacts (e.g., muscle twitch).
ms before stimulus
Furthermore,
replicated using a Biosemi ActiveTwo™ [16] 128
channel EEG system for one subject. This was
important because the Biosemi, unlike the MRplus,
is an unshielded system. It was possible, therefore,
that electrical noise from the Stewart platform
would interfere with the data from the unshielded
system. The data acquired with the Biosemi were
digitized at 2048Hz, with a hardware lowpass filter
at 400Hz applied during data acquisition. They
were downsampled to 256Hz. The data from the Pz
channel was epoched from 1000ms pre-stimulus to
2000ms post-stimulus, baseline-corrected from
-500ms to stimulus onset, and then filtered between
2 and 45Hz. For the BrainAmp and Biosemi data,
poor quality epochs were identified and rejected,
using an automated method based on epoch
amplitude, deviation and variance.
Following this step, specifically for the Biosemi
data, Independent Component Analysis (ICA) was
performed, and noisy and artifactual components
were identified and rejected by an automated
method based on a component’s maximum
component standard deviation, correlation with the
EOG channels, contribution to the EEG waveform,
spectral power and spectral slope. ICA was not
feasible for the BrainAmp data due to the low
number of channels, which would not allow for a
good approximation of statistical independence.

these experiments were later

Figure 1. Setup and stimuli. a Graphical representation of the
Stewart motion platform. b Experimental setup.

C. Visual oddball paradigm
The visual oddball paradigm consisted of different
sized circles presented for 60 ms on the screen,
separated by an inter-stimulus interval of 2000 ms,
for 205 trials in a pseudorandom order. Frequent
non-target (80%) and infrequent target (20%)
circles differed in size and subtended 2 degrees and
6 degrees, respectively. Subjects were instructed to
press a button as quickly as possible following a
target (large) circle. The visuals were displayed
using Presentation™.
D. Motion Stimuli
There were four different motion stimuli: static,
idle, slow and fast. In the static condition the
platform was turned off. In the idle condition the
platform was raised to its neutral position then
remained idle in this position. In the slow and fast
conditions the platform was moved sinusoidally in
a fore-aft motion at a frequency of 0.5 Hz for 230
b
586
Authorized licensed use limited to: TRINITY COLLEGE LIBRARY DUBLIN. Downloaded on August 28, 2009 at 04:28 from IEEE Xplore. Restrictions apply.

Page 3

cycles at different maximum accelerations of
0.1mG and 0.5mG, respectively. The motions were
controlled by Virtools™.
E. Experiment procedure
Participants performed the visual oddball task
during all four motion conditions (static, idle, slow
and fast). In the idle, slow and fast conditions the
motion platform was raised to the neutral position
and motion was started before the beginning of the
oddball paradigm. The timing of the onset of the
visual stimulus was pseudorandom to ensure it was
not locked to motion stimulus.

III.
RESULTS

TABLE 1
P300 CORRELATION COEFFICIENTS BETWEEN
MOTION CONDITIONS
Idle Slow
Static 0.9766 0.9394
Idle 0.9326
Slow


Figure 2 shows the average ERP data (n=5) for
each motion condition. A robust and characteristic
peak around 300ms can be seen in all the target
(grey) ERPs. Table 1 shows the correlation
coefficients during the interval -100ms to 600ms of
the ERPs, which can all be seen to be above 0.93.
For display purposes, the ERP data in all figures
have been subject to a 13Hz low-pass filter.




Figure 3 shows the spectra for each of the motion
conditions for a representative subject (Subject 2).
The spectra were subject to a 20-point moving
average smoothing filter for display purposes, as
there were a large number of data points in the
resultant FFT. These spectra produce a covariance
matrix with a mean of 5.9823±0.5226, and a
variance of 0.2731.



Figure 4 shows a P3 recorded with the Biosemi
system.
On inspecting epochs, we found there was no
rejection required for the BrainAmp data. The
automated rejection tool selected 5 out of 5536
epochs across all 5 subjects, from a static, slow and
fast dataset for the Biosemi data.
IV. DISCUSSION
We found that there was minimal interference to
the EEG signal during motion of the participants on
the platform, using both the shielded MRplus
system and the unshielded Biosemi system. A
reliable P3 ERP was evoked in all conditions, as
demonstrated by the peak around 300ms in the
target ERP (black), which was not present in the
standard ERP (grey). Furthermore, as Table 1
shows, the correlation between the static (motors-
off) condition and the other 3 conditions is similar
to the correlation between
conditions, indicating that the recorded data are
unaffected by the platform’s state. This is also
reflected on examination of the rejected epochs,
which show no electromagnetic artifacts caused by
the motors-on
Figure 3. Spectra of each motion condition for Subject 2.
Figure 2. Averaged ERP data (n=5) recorded under
all motion conditions from Pz acquired using
BrainAmp MRplus. Both target (black) and
standard (grey) averages are shown.
Static Idle
Slow Fast
Fast
0.9682
0.9687
0.9488
Figure 4. ERP data (n=1) recorded using the Biosemi ActiveTwo
system during the Idle condition. The target (black) and standard
(grey) are shown.
587
Authorized licensed use limited to: TRINITY COLLEGE LIBRARY DUBLIN. Downloaded on August 28, 2009 at 04:28 from IEEE Xplore. Restrictions apply.

Page 4

the Stewart platform, and negligible difference in
rejection rates across different motion conditions.
More evidence can be seen by inspecting Figure 3,
which shows the spectra of each of Subject 2’s
sessions, overlaid. There are no large deviations
from the overall trend, as can be seen by visual
inspection, and shown in the small standard
deviation and variance in the covariances between
the spectra. These properties are seen in all
subjects.
Figure 4 shows a sample plot from the Biosemi P3
recordings. The peak at 300ms is present but not as
well-defined as those in Fig. 2; as Fig. 4 depicts
only a single subject, more testing will be required
to determine whether this is due to inter-subject
variability or a lower quality signal in the Biosemi.

These results are of interest because we have
shown that EEG can be acquired during linear self-
motion, which has never been demonstrated before
in humans. The methodology described in this
study may provide a basis for investigating the
functioning of the vestibular system. In particular,
the combined use of a Stewart platform and EEG
recording could facilitate the investigation of age-
related disorders of the vestibular system, and
vestibular functional decline. This method may also
be useful for further investigating MSI in the
vestibular system. Future
investigate the relationship between the visual and
vestibular system via a combined Visual-Vestibular
P3 experiment.

experiments will
V.
REFERENCES
[1] H. Straka, N. Dieringer, Basic organization
principles of the VOR: lessons from frogs, Progress
in Neurobiology, Vol. 73(4), pp. 259-309, July
2004.
[2] F. B. Horak, G. M. Earhart V. Dietz, Postural
responses to combinations of head and body
displacements: vestibular-somatosensory
interactions, Exp Brain Res, Vol. 141, pp. 410-414,
2001.
[3] J. O'Loughlin et al. Incidence of and risk factors
for falls and injurious falls among the community-
dwelling elderly, American journal of
epidemiology, Vol. 137, pp. 342-54, 1993.
[4] M. E. Tinetti, M. Speechley, Prevention of falls
among the elderly, New England journal of
medicine, Vol. 320, pp. 1055-1059, 1989.
[5] S. R. Lord, D. McLean, G.
Strathers, Physiological factors associated with
injurious falls in older people living in the
community, Gerontology, Vol. 38, pp. 338-346,
1992.
[6] World Health Organization, Assessment of
fracture risk and its application to screening for
postmenopausal osteoporosis, Technical report
series 843, Geneva, 1994.

[7] G. Kovacs, M. Raabe, and M. W. Greenlee, The
fMRI correlates of visually induced self-motion in
depth, Cereberal Cortex, Vol. 18, pp. 1779-1787,
2008.
[8] T. Brandt, P. Bartenstein, A. Janek, and
M Dieterich. Reciprocal inhibitory visual-
vestibular interaction, Reciprocal inhibitory visual-
vestibular interaction. Visual motion stimulation
deactivates the parieto-insular vestibular cortex,
BRAIN, Vol. 121, pp. 1749-1758, 1998.
[9] D. Stewart, A Platform with Six Degrees of
Freedom, UK Institution of Mechanical Engineers
Proceedings, Vol. 180, Pt 1, No 15, 1965-66.
[10] D. Schneider, C. Kolchev, L. Constantinescu,
C.-F. Claussen, Vestibular Evoked Potentials and
Brain Electrical Activity Mapping - A Test of
Vestibular Function - A Review, Int. Tinnitus
Journal, Vol. 2, pp. 27-43, 1996.
[11] J. D. Hood, A. Kayan, Observations upon the
evoked responses to natural vestibular stimulation,
Electroencephalography and Clinical
Neurophysiology, Vol. 62, pp. 266-276, 1985.
[12] S. Sutton, M. Braren, J. Zubin, E. R John,
Evoked-Potential Correlates of Stimulus
Uncertainty, Science, 150(3700), pp. 1187-1188,
1965.
[13] M. von der Heyde, A Distributed Virtual
Reality System for Spatial Updating: Concepts,
Implementation, and Experiments, PhD thesis,
Universität Bielefeld - Technische Fakultät, 2001.
[14] http://www.brainproducts.com/ - Brain
Products GmbH.
[15] A. Delorme, S. Makeig, EEGLAB: an open
source toolbox for analysis of single-trial EEG
dynamics including independent component
analysis, J. Neurosci Methods, Vol. 134, pp. 9-21,
2004.
[16] http://www.biosemi.com/ - Biosemi company

588
Authorized licensed use limited to: TRINITY COLLEGE LIBRARY DUBLIN. Downloaded on August 28, 2009 at 04:28 from IEEE Xplore. Restrictions apply.