A 12-channel, real-time near-infrared spectroscopy instrument for brain-computer interface applications.

Abstract

A continuous wave near-infrared spectroscopy (NIRS) instrument for brain-computer interface (BCI) applications is presented. In the literature, experiments have been carried out on subjects with such motor degenerative diseases as amyotrophic lateral sclerosis, which have demonstrated the suitability of NIRS to access intentional functional activity, which could be used in a BCI as a communication aid. Specifically, a real-time, multiple channel NIRS tool is needed to realise access to even a few different mental states, for reasonable baud rates. The 12-channel instrument described here has a spatial resolution of 30 mm, employing a flexible software demodulation scheme. Temporal resolution of approximately 100 ms is maintained since typical topographic imaging is not needed, since we are only interested in exploiting the vascular response for BCI control. A simple experiment demonstrates the ability of the system to report on haemodynamics during single trial mental arithmetic tasks. Multiple trial averaging is not required.

Full text

Page 1

30th Annual International IEEE EMBS Conference
Vancouver, British Columbia, Canada, August 20-24, 2008

A 12-Channel, real-time near-infrared spectroscopy instrument for
brain-computer interface applications
C. Soraghan, F. Matthews, C. Markham, B.A. Pearlmutter, R. O’Neill, T.E. Ward


Abstract— A continuous wave near-infrared spectroscopy
(NIRS) instrument for brain-computer interface (BCI)
applications is presented. In the literature, experiments have
been carried out on subjects with such motor degenerative
diseases as amyotrophic lateral sclerosis, which have
demonstrated the suitability of NIRS to access intentional
functional activity, which could be used in a BCI as a
communication aid. Specifically, a real-time, multiple channel
NIRS tool is needed to realise access to even a few different
mental states, for reasonable baud rates. The 12-channel
instrument described here has a spatial resolution of 30mm,
employing a flexible software demodulation scheme. Temporal
resolution of ~100ms is maintained since typical topographic
imaging is not needed, since we are only interested in exploiting
the vascular response for BCI control. A simple experiment
demonstrates the ability of the system to report on
haemodynamics during single trial mental arithmetic tasks.
Multiple trial averaging is not required.
I. INTRODUCTION

U
NLOCKING the brain and understanding better its
capabilities, functioning, and pathologies is at the
forefront of biological science research currently. In an
attempt to tap into the brain’s numerous simultaneous
processes during particular mental operations, some success
has been attained in exploiting these processes for
controlling augmentative devices to improve the quality of
life of those with seriously impaired functional expression
[1].
It has been advised, at least for certain subjects, that
implementation of non-invasive modalities would be a more
prudent approach to first tackle the issue and demand for a
brain-computer interface (BCI) by an individual [2]. Several
groups have dedicated research to the area, many opting for
the more utilised agent of electrophysiology of the brain,
namely surface electroencephalography (EEG) [3]. Other
BCI modalities have been explored such as functional
magnetic resonance imaging
magnetoencephalography (MEG). However these are less
(fMRI) and

Manuscript received April 7, 2008. This work is supported by Science
Foundation Ireland (SFI) grant number SFI/05/RFP/ENG0089.
All authors are with the National University of Ireland, Maynooth
(corresponding author email: christopher.j.soraghan@nuim.ie).
Authors Departments (some authors are with 2 departments): Dept. of
Experimental Physics (C. Soraghan & R. O’Neill), Dept. of Computer
Science (C. Soraghan, F. Matthews, & C. Markham), Dept. of Electronic
Engineering (T.E. Ward), Hamilton Institute (B.A. Pearlmutter, & F.
Matthews)

practical due to their size and the ambient requirements.
Another method of brain imaging, termed near-infrared
spectroscopy (NIRS), has been developed over the last 3-4
decades and its permissibility for use in a BCI has been
suggested by several research groups now pursuing the
optical BCI (OBCI) modality [4]-[9]. The ability of NIRS to
inspect brain state in amyotrophic lateral sclerosis (ALS) has
also been demonstrated recently [4], [10]. It has also been
reported that an OBCI may be a suitable tool in stroke
rehabilitation [11]. Indeed it has also been mooted that an
OBCI may have applications as a complementary
interrogatory method allying with, in particular, EEG [12],
[13]. For instance, the supplementary direct localisable
abilities of NIRS could be harnessed alongside the high
temporal resolution of EEG.
In this paper we wish to describe a dedicated 12-channel
NIRS instrument for BCI applications. The device is capable
of real-time acquisition of functional activity that is relayed
in vascular haemodynamic concentration changes. The
instrument is supported by a LabView® interface platform
that synchronises, displays, and processes haemodynamic
data.
Current embodiments of NIRS-BCIs are limited in terms
of online real-time capability or in the case of real-time
applicability, they are limited in the area of tissue they can
interrogate. Our current device has improved SNR and
localisation by having multiple channels as well as utilising
multiple wavelength sources for increased spatial resolution.
In addition, tissue interrogation paradigms are very flexible
since independent sources and detector-fibres are not
constrained to fixed optode arrangements on the head. Thus,
custom-made holders are/can be made for experiment-
specific irradiation patterns. Consequently, instrumentation
can be minimized for subject comfort.
II. OPTICAL BCI BACKGROUND
A. Optical Brain-Computer Interface
A brain-computer interface is simply a device that
presents a user with a viable communication aid for
translating/transducing thoughts (e.g. cognitive, motor
control) into observable gestures, for example, moving a
cursor on a computer screen. An optical NIRS-BCI is a
device that exploits the change in optical properties of focal
cortical brain tissue during functional activation for such
communication.
978-1-4244-1815-2/08/$25.00 ©2008 IEEE.5648
Authorized licensed use limited to: The Library NUI Maynooth. Downloaded on July 27, 2009 at 11:04 from IEEE Xplore. Restrictions apply.

Page 2

B. Near-Infrared Spectroscopy
NIRS is essentially a tool whereby matter is penetrated by
NIR radiation and the resultant absorption and scattering
effects are observed so as to deduce the matter’s
composition.
Assumptions are made about the absorbing molecules
(chromophores) involved in brain imaging. In cortical tissue
and other superficial cranial layers some chromophores can
be considered constant absorbers. The main constant
absorbers of NIR light that are found in brain tissue are
melanin, water, lipids, and cerebrospinal fluid (CSF). At
normal concentrations in blood, plasma can similarly be
treated as a colourless liquid in comparison to haemoglobin
absorption. Haemoglobin and various respiratory enzymes,
in particular cytochrome-c-oxidase (ccOXY), are the
varying chromophores of interest which indicate cerebral
blood oxygenation and tissue oxygenation [14].
With any mental action performed, neuronal responses are
manifested in various regions of the brain. With this,
neighbouring vascular responses ensue (neurovascular
coupling) whereby a combination of increases in cerebral
blood flow (CBF), cerebral blood volume (CBV), and
cerebral metabolic rate of oxygen consumption (CMRO2)
lead to an overall localised increase in oxy-haemoglobin
(HbO) and a relatively smaller localised decrease in deoxy-
haemoglobin (HbR). Again, varying haemoglobin is the
main marker of tissue oxygenation dynamics. A modified
version of the Beer Lambert Law can be used here to
determine HbO and HbR concentration changes:



(1)



where
HbO
c
Δ
and
HbR
c
Δ
are determined concentration
A
Δ
changes in HbO and HbR respectively;
recorded wavelength specific light intensities; B760 and B880
(constants) are the product of the direct inter-optode spacing
and the tabulated dependent differential pathlength factors. α
represents the specific tabulated extinction coefficients for
either HbO or HbR at each specific wavelength [14].

and
880760
A
Δ
are
III. SYSTEM DESCRIPTION
A. Overview
The 12–channel (24 LEDs, 3 detectors) continuous wave
NIRS (cwNIRS) system at NUI Maynooth has a new
interfacing and processing paradigm that makes use of a
software based demodulation scheme. This is to replace the
dual channel system by Coyle [13] where they envisaged an
extension to the capabilities of the first prototype, in order to
interrogate a larger surface area of the cerebral cortex to
ultimately increase the throughput of the optical BCI [15].
Instead of the analogue based lock-in detection techniques
used until now [11]-[13], [15]-[16], a LabView® (National
Instruments graphic programming language) centered
workstation controls all aspects of the instrument from
driving LEDs, performing spectral analysis to visualisation.
All processing, feedback, and synchronisation is dealt with
by various LabView® virtual instruments.
The main components of the cwNIRS device are the
sources and driving electronics, detectors, coupling optics,
mechanical harnessing, and the data acquisition and
processing unit. These are illustrated in Fig. 1.
















Fig. 1: Data flow diagram of the cwNIRS tool for accessing brain state.
Abbreviations are expanded in the main text.

The remainder of Section III is used to describe the
instrument in terms of its optical components, electronic
components, and software interface. Section IV reports on
the haemodynamic signals obtained by the device along with
a brief description of their significance. Finally, Section V
concludes the paper with a discussion.
PC
LabView

APD1

APD3

APD2
Controller
(PXI-Chassis 1033)

AO

ADC1

ADC2
T
I
S
S
U
E
LED Driver
12 triple
wavelength LEDs
Fibre
Bundles
24-Sinusoidal
biased signals
760880
760880 880760
760880
760880 880760
880760
880760
880760
880760
;
HbRHbR
HbO
HbRHbO
A
B
HbR HbO
A
B
HbOHbO
HbR
HbOHbRHbOHbR
AA
BB
c
c
αα
αα
Δ
αα
Δ
αα
αααα
ΔΔ
−
Δ=
−
−
Δ=
−
B. Optical Components
1) Light Source
Custom-made,
wavelength
APT0101) were integrated into the device. When applied
correctly, they allow sufficient photons to penetrate the
cerebral cortex and reflect back after intercepting the
localised cortical haemodynamic traffic. Only two of the
three dyes per LED are currently being used, with 760nm to
resolve HbR and 880nm for resolving HbO (800nm dye is
not being used). With 30nm and 80nm spectral bandwidth at
50% maximum (full width half maximum), chromophore
crosstalk is avoided, at least optically. An 8° beam angle
with perpendicular application of the LEDs to the scalp
augments the irradiance reaching the cortical surface. The
co-location and lensing of the dyes in each LED package
helps improve precision during measurements.
high power (5mW-17mW),
(Opto
triple
Corp: NIR-AlGaAs-LEDs Diode
5649
Authorized licensed use limited to: The Library NUI Maynooth. Downloaded on July 27, 2009 at 11:04 from IEEE Xplore. Restrictions apply.

Page 3

2) Detector
The avalanche photodiode (APD) detector chosen to
collect the back-reflected photons (Hamamatsu, C5460-01)
is/has being used by other NIRS groups [17], [18], which
has a high sensitivity (0.005nW) and excellent noise-
equivalent power (NEP) of 20fW/Hz1/2. With high
attenuation (7-9 orders of magnitude) in biological tissue,
NIR light is said to accomplish traversing ~8cm of tissue
from the source before being fully absorbed [19]. The APD
with an active area of 3mm requires glass fibre optic bundles
(Edmund Optics, NT40-644) of similar diameter and a large
acceptance angle to guide the photons from the scalp to the
detector. The APD requires only a ±12V regulated DC
supply to generate the high internal voltage (~200V)
necessary for the avalanche-multiplication process.

3) Mechanical Optode Holder
A major contributor to the overall signal integrity is the
mechanical restraining of the sources and fibre optics to the
subjects scalp, all the while maintaining subject comfort.
Many configurations of optode coupling have been
implemented including a dedicated mechanical mounting
system constructed by Coyle [20]. However, in its
development and to support a multi-channel system, we
fashioned brass tubing in felt and hot melt glue to house the
sources and detectors in a 1 detector, 6 source ‘optet1’. The
6 LEDs encircle the detector with a 30mm radial spacing
(See Fig. 2). This allows quick localising of more active
channels, which, once identified, will then allow for
replacement of the optet with 1-3 LEDs and a single
detector, for the final BCI design. In this way the number of
independent BCI channels can be increased to augment the
OBCI’s throughput, inspecting, with minimal sources and
detectors, only the areas of maximal activity for each
cortical function (e.g. cognitive, motor, visual).
C. Electronic Components
1) LED Driver
An LED Driver was constructed to provide a stable
current-to-intensity source to drive the 12 triple wavelength
LEDs. The overall system is scalable to accept many more
LEDs, the only constraint being the modulation bandwidth
available to separate the sources. All LEDs are modulated
(sinusoidal) in the range 1kHz-20kHz, each driven at a
maximum of 100mA (~160mW). A genetic algorithm based
application was written for selecting frequencies with
minimal overlap and harmonic distortion in the 19kHz
range.

2) Data Acquisition and Processing Unit
The entire cwNIRS system is integrated within a single
workstation running LabView®. This interfaces with
devices for signal acquisition (NI-PXI4462), and waveform
generation (NI-PXI6723) for the LED driver. The PXI4462
is a 24-bit, 4-channel, 204kHz simultaneous sampling rate
analogue-to-digital (ADC) converter. The PXI6723 is a
multifunction digital-to-analogue converter (DAC) with 32-
analog output channels (AO). Two ADCs and one DAC are
housed in a PXI-1033 chassis, which is used to buffer and
synchronise data transfer to the workstation via a PCI-
Express card. All cards are operated and synchronised to the
same 10MHz master clock via a trigger line on the chassis
backpane, ensuring robust data timing. The PXI chassis has
a further 2 empty slots for either ADC or DAC to allow for
system scaling of sources/detectors.

1 We term a single source and detector an ‘optode’. We term an optode
with multiple spatially disparate sources an ‘optet’.
D. Software Interface
Source selection by demodulation is performed in
software as inspired by the work of Everdell [17]. Using
Fast-Fourier Transform (FFT) demodulation, the amplitude
of each dye’s carrier signal is extracted from the bulk data,
thus separating the 12 dye’s light intensities (6 LEDs, 2 dye
per LED) reaching each detector. The signal processing
pipeline continues within LabView®, where the data is
downsampled to 20Hz. Next, filtering and calculation of
haemoglobin concentration
Classification of BCI activation events, and user feedback
with BCI device control can subsequently be carried out in
real-time. During an experiment, the user is provided with
visual or auditory prompts for, for example, mental
arithmetic tasks. The resultant BCI classification of trials is
then either reported to the user in training, or used to
perform some communication control. In addition, success
rates are fed back to the user.
changes is performed.
IV. SYSTEM VALIDATION
A short experiment was carried out to verify that the
system could indeed report on the cerebral haemodynamics
of the user. In future experiments these will be exploited to
classify intentional functional activity, for application in a
BCI to provide communication aid.
Two subjects (mean age 25 years old) took part in the
experiment. Both subjects had previous experience with an
optical BCI. All 12 LEDs, but only 2 detectors, were placed
on the frontal lobe in two optets of 6 sources and 1 detector
each. They encapsulated cortical tissue around F7-Fp1 and
F8-Fp2 of the 10-20 electrode placement system (see Fig. 2).
The subjects were asked to perform mental arithmetic
exercises during synchronous predetermined trials which
simply entailed successively
subtraction the minuend is the result of the previous
subtraction) two whole randomised numbers (the minuend
was a 4-digit number, the subtrahend a 2-digit number
between 16 and 22). The initial rest period was fixed at 30s
and the remaining rest periods were randomised between
15s-30s, in an attempt to curtail habituation, stimulus
expectation, and low blood oscillation (Mayer Wave)
synchronisation. Stimulus periods were fixed at 15s. Each
subtracting (for each
5650
Authorized licensed use limited to: The Library NUI Maynooth. Downloaded on July 27, 2009 at 11:04 from IEEE Xplore. Restrictions apply.

Page 4

subject performed 2 sets of 10 trials within the same session.
Both subjects reported similar results. Fig. 2 shows the
non-averaged response from subject 1 on channel 3 after
band pass filtering (0.05Hz-0.2Hz). This non-averaged
analysis is illustrated to report the possibility of an OBCI to
determine single trial functional activations, in real-time.
After locating the channels with maximum activation, the
number of LEDs per optet would be typically be reduced,
perhaps to 1-3 LEDs. Differences in neighbouring channels
were also observed demonstrating that the cwNIRS system
reports spatially varying haemodynamics.
I. DISCUSSION
A 12-channel cwNIRS instrument has been constructed
and described here for use in an optical BCI. Since the
device uses single source-detector pairing the spatial
resolution is ≤ inter-optode spacing of ~30mm with detector
sensitivity of 0.005nW. Irradiance of ~5mW-17mW
provides adequate cerebral penetration, and the data
acquisition (24-bit 204kS/s ADC) allows for sensitive
analysis of haemodynamics, as reported in Fig. 2. A 10MHz
shared clock ensures synchronisation between data
collection and generation, also allowing for accurate data
time-stamping. Software demodulation schemes allow for
ease of expansion of the device. The flexible LabView®
interface allows for rapid prototyping of BCI paradigms,
also allowing speedy testing of signal processing techniques.
The system can also be used with HomER (see [12]) to
produce offline topographic mapping of interrogated
cerebral tissue.
Future work will consist of running a series of
experiments to test the
computational overhead speed) of an optical BCI using
cognitive/other mental tasks.
performance (baud rate,
REFERENCES
[1] N. Birbaumer, A. Kubler, N. Ghanayim, T. Hinterberger, J.
Perelmouter, J. Kaiser, I. Iversen, B. Kotchoubey, N. Neumann, and
H. Flor, "The thought translation device (TTD) for completely
paralyzed patients," IEEE Transactions on Rehabilitation Engineering,
vol. 8, pp. 190-193, Jun 2000.
[2] R. A. Bakay, "Limits of brain-computer interface. Case report,"
Neurosurg Focus vol. 20, p. E6, 2006.
[3] J. R. Wolpaw, D. J. McFarland, and T. M. Vaughan, "Brain-computer
interface research at the Wadsworth Center," IEEE Trans Rehabil Eng
vol. 8, pp. 222-6, Jun 2000.
[4] M. Naito, Y. Michioka, K. Ozawa, Y. Ito, M. Kiguchi, and T.
Kanazawa, "A communication means for totally locked-in ALS
patients based on changes in cerebral blood volume measured with
near-infrared light," IEICE Transactions on Information and Systems
vol. E90d, pp. 1028-1037, Jul 2007.
[5] R. Sitaram, H. Zhang, C. Guan, M. Thulasidas, Y. Hoshi, A. Ishikawa,
K. Shimizu, and N. Birbaumer, "Temporal classification of
multichannel near-infrared spectroscopy signals of motor imagery for
developing a brain-computer interface," Neuroimage Dec 28 2006.
[6] G. Bauernfeind, R. Leeb, S. C. Wriessnegger, and G. Pfurtscheller,
"Development, set-up and first results for a one-channel near-infrared
spectroscopy system," Biomed Tech (Berl), vol. 53, pp. 36-43, 2008.
[7] H. Ayaz, "Detecting cognitive activity related hemodynamic signal for
brain computer interface using functional near infrared spectroscopy,"
in IEEE EMBS Conference on Neural Engineering Kohala Coast,
Hawaii, USA, 2007.
[8] H. Ogata, T. Mukai, and T. Yagi, "A Study on the Frontal Cortex in
Cognitive Tasks using Near-Infrared Spectroscopy," Conf Proc IEEE
Eng Med Biol Soc vol. 1, pp. 4731-4, 2007.
[9] J. Kohlenberg and T. Chau, "Detecting Controlled Signals in the
Human Brain by Near Infrared Spectroscopy," in Proceedings of the
28th IEEE EMBS Annual International Conference New York City,
USA, 2006.
[10] Y. Fuchino, M. Nagao, T. Katura, M. Bando, M. Naito, A. Maki, K.
Nakamura, H. Hayashi, H. Koizumi, and T. Yoro, "High cognitive
function of an ALS patient in the totally locked-in state," Neurosci
Lett, vol. 435, pp. 85-9, Apr 18 2008.
[11] T. E. Ward, C. J. Soraghan, F. Matthews, and C. Markham, "A
Concept for Extending the Applicability of Constraint-Induced
Movement Therapy through Motor Cortex Activity Feedback Using a
Neural Prosthesis," Comput Intell Neurosci, p. 51363, 2007.
[12] F. Matthews, B. A. Pearlmutter, T. E. Ward, C. Soraghan, and C.
Markham, "Hemodynamics for braincomputer interfaces," IEEE
Signal Processing Magazine, vol. 25, pp. 87-94, Jan 2008.
[13] S. Coyle, "Near-Infrared Spectroscopy for Brain Computer
Interfacing," in Engineering vol. PhD Thesis Kildare: Maynooth,
2005, p. 222.
[14] M. Cope, "The application of NIRS to non-invasive monitoring of
cerebral oxygenation in the newborn infant," in Medical Physics and
Bioengineering vol. PhD London: London, 1991, p. 342.
[15] S. M. Coyle, T. E. Ward, and C. M. Markham, "Brain-computer
interface using a simplified functional near-infrared spectroscopy
system," Journal of Neural Engineering vol. 4, pp. 219-26, Sep 2007.
[16] C. Soraghan, F. Matthews, D. Kelly, T. Ward, C. Markham, B.
Pearlmutter, and R. O'Neill, "Viability of a dual channel optical BCI
for use in a gaming environment," in CGAMES Dublin Institute of
Technology, Dublin, Ireland, 2006.
[17] N. L. Everdell, A. P. Gibson, I. D. C. Tullis, T. Vaithianathan, J. C.
Hebden, and D. T. Delpy, "A frequency multiplexed near-infrared
topography system for imaging functional activation in the brain,"
Review of Scientific Instruments vol. 76, pp. -, Sep 2005.
[18] D. K. Joseph, T. J. Huppert, M. A. Franceschini, and D. A. Boas,
"Diffuse optical tomography system to image brain activation with
improved spatial resolution and validation with functional magnetic
resonance imaging," Appl Opt vol. 45, pp. 8142-51, Nov 1 2006.
[19] C. E. Elwell, A practical users guide to nrear-infrared spectrocopy.
London, UK, 1995.
[20] S. Coyle, C. Markham, W. Lanigan, and T. Ward, "A Mechanical
Mounting System for Functional Near-Infrared Spectroscopy Brain
Imaging Studies," in Opto-Ireland Ireland, 2005.
Fig. 2: Ten trials of mental arithmetic tasks from subject 1 on channel 3. During the stimulus periods (shaded regions), activity in the left frontal cortex is
indicated by increased haemodynamics, in particular, an increase in HbO and a decrease in HbR. The optet arrangement used is shown on the left.
1 2 3 4 5 6
APD fibre
LEDs


0 50 100150200 250300350
-10
-5
0
5
x 10
-4
Time (s)
Conc. Changes
(mM.cm)
HbO
HbR
5651
Authorized licensed use limited to: The Library NUI Maynooth. Downloaded on July 27, 2009 at 11:04 from IEEE Xplore. Restrictions apply.