186 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO.
2, APRIL 2013
Wireless Recording Systems: From Noninvasive
EEG-NIRS to Invasive EEG Devices
Mohamad Sawan, Fellow, IEEE, Muhammad T. Salam, Jérôme Le Lan, Amal Kassab, Sébastien Gélinas,
Phetsamone Vannasing, Frédéric Lesage, Maryse Lassonde, and Dang K. Nguyen
Abstract—In this paper, we present the design and imple-
mentation of a wireless wearable electronic system dedicated to
remote data recording for brain monitoring. The reported wire-
less recording system is used for a) simultaneous near-infrared
spectrometry (NIRS) and scalp electro-encephalography (EEG)
for noninvasive monitoring and b) intracerebral EEG (icEEG)
for invasive monitoring. Bluetooth and dual radio links were
introduced for these recordings. The Bluetooth-based device was
embedded in a noninvasive multichannel EEG-NIRS system for
easy portability and long-term monitoring. On the other hand, the
32-channel implantable recording device offers 24-bit resolution,
tunable features, and a sampling frequency up to 2 kHz per
channel. The analog front-end preamplifier presents low input-re-
ferred noise of 5 and a signal-to-noise ratio of 112 dB.
The communication link is implemented using a dual-band radio
frequency transceiver offering a half-duplex 800 kb/s data rate,
16.5 mW power consumption and less than post-correction
Bit-Error Rate (BER). The designed system can be accessed and
controlled by a computer with a user-friendly graphical interface.
The proposed wireless implantable recording device was tested in
vitro using real icEEG signals from two patients with refractory
epilepsy. The wirelessly recorded signals were compared to the
original signals recorded using wired-connection, and measured
normalized root-mean square deviation was under 2%.
Index Terms—Bluetooth and dual radio links, brain imaging,
electroencephalography (EEG), embedded systems, intracerebral
EEG, near-infrared spectrometry, refractory epilepsy, wearable
bioelectronics, wireless data recording.
I. INTRODUCTION
H UMAN brain signal recording is important for both re-
search purposes and assessment of various neurological
disorders [1]–[3]. For example, EEG is the current method of
Manuscript received October 04, 2012; revised December 28, 2012; accepted
March 18, 2013. Date of publication April 25, 2013; date of current version May
03, 2013. This work was supported in part by the Canadian Institutes of Health
Research (CIHR), the Institute of Circulatory and Respiratory Health (ICRH),
the Heart and Stroke Foundation of Canada (HSFC), the Canada Research Chair
on Smart Medical Devices, and ReSMiQ. This paper was recommended by As-
sociate Editor S. Carrara.
M. Sawan, M. T. Salam, J. Le Lan, A. Kassab, and S. Gélinas are
with Polystim Neurotechnologies Laboratory, Electrical Engineering
Department, Polytechnique Montréal, QC H3T1J4, Canada (e-mail: mo-
hamad.sawan@polymtl.ca).
P. Vannasing and M. Lassonde are with Sainte-Justine Hospital, Centre Hos-
pitalier Universitaire Mère-Enfant, Montréal, QC H3T 1C5, Canada.
F. Lesage is with the Optical and Molecular Imaging Laboratory, Polytech-
nique Montréal, QC H3C 3A7, Canada.
D. K. Nguyen is with the Neurology Service, Department of Medicine, Notre-
Dame Hospital, Centre Hospitalier de l’Université de Montréal, QC H3T 1J4,
Canada.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TBCAS.2013.2255595
1932-4545/$31.00 © 2013 IEEE
choice to visualize abnormal epileptiform discharges in patients
with epilepsy [4]. Continuous EEG monitoring is commonly
used for the diagnosis and monitoring of convulsive or non-con-
vulsive status epilepticus and assessment of ongoing therapy for
the treatment of seizures in such patients [5], [6]. Some patients
may benefit from epilepsy surgery if the epileptogenic zone
(EZ) can be identified and resected without harm. There are two
steps for epileptogenic zone localization: a) noninvasive, and b)
invasive brain signal monitoring. The noninvasive monitoring
can roughly estimate seizure activation region. Moreover, due to
the limited spatial or temporal resolution of currently available
noninvasive localization techniques, accurate delineation of the
EZ may sometimes be arduous, particularly with non-lesional
refractory epilepsy. Therefore, following noninvasive studies,
an invasive EEG recordings may be necessary prior to resective
surgery. Near-infrared spectroscopy (NIRS), which can mon-
itor local changes in cerebral blood volume and oxygenation
noninvasively [7], [8], is used both in research to identify areas
involved in cognitive tasks [9], [10] and in clinical settings to
monitor cerebral oxygenation during cardiac or carotid surgeries
[11], [12]. Long-term invasive monitoring, over 2–3 weeks, are
performed in epilepsy centers to record seizures in order to de-
lineate the area of seizure onset for curative resection [3], [5],
[13] and low-noise preamplifiers would be beneficiary for this
application [14]–[16].
During these monitoring sessions, patients are connected to
relatively bulky machines in a small controlled environment.
Wireless portable devices could provide important benefits [3].
For example, a portable wireless EEG-NIRS system could allow
researchers to study brain activity during dynamic tasks such as
gait rehabilitation following a stroke [11]. Similarly, a wireless
invasive EEG system ([17], [18]) could reduce the risk of in-
fections associated with the use of intracranial electrodes and
their exit cables connected to external amplifiers. One could also
envision implant a small recording device chronically which
would wirelessly transmit a stream of EEG data allowing online
seizure detection [19] and triggering a warning/alarm system or
focal treatment (thermal [20], drug [21] or stimulation [22]). For
the latter purposes, 2 kHz sampling rate is required for early de-
tection of low-voltage fast oscillations seen at seizure onset [23],
[24].
In this paper, we present two wireless brain signal-recording
systems, one noninvasive [Fig. 1(a)] and the other invasive
for cerebral monitoring [Fig. 1(b)]. The noninvasive one is a
portable wireless EEG-NIRS system, which has 8 channels
for scalp EEG and NIRS recording. Our previous prototype
[12] did not have wireless data-communication. This proposed
system is an upgraded, portable, battery-powered device to
SAWAN et al.: WIRELESS RECORDING SYSTEMS: FROM NONINVASIVE EEG-NIRS TO INVASIVE EEG DEVICES
Fig. 1. Device configuration. (a) Noninvasive EEG-NIRS system. (b) Invasive
wireless icEEG system.
minimize patient discomfort for longer periods of monitoring.
Moreover, the proposed dual 8-channel recording device has
high sensitivity to acquire good quality optical and electrical
signals with the presence of scalp hair.
The invasive system represents a new implantable wireless
icEEG recording device designed for presurgical evaluation.
The goal is to reduce the risk of infections and improve patient
mobility during the 2 to 3 weeks study. The proposed device is
implanted on the skull under the skin and is used to transmit the
icEEG recordings through the skin to a remote base station for
continuous monitoring. It features 32-channel of wireless raw
recordings with 24-bit resolution, offers tuneability and ease of
control of the implant remotely. Also, the architecture of the
proposed system is scalable, which allows increasing channels
with less effort. Although such systems could be used for a va-
riety of purposes, we focus in this paper on their potential ap-
plication to the field of epilepsy.
II. PROPOSED SYSTEMS
Patients with epilepsy may sometimes require noninvasive
and/or invasive EEG studies for a variety of reasons: diagnosis
of episodic events, localization of seizure focus, monitoring
of seizure activity, etc. The proposed wireless techniques and
system implementation are described as follows.
A. Noninvasive Studies
The achieved noninvasive prototype is a combined
EEG-NIRS portable brain imaging system, implemented
for long term monitoring. Fig. 2 illustrates a simplified func-
tional diagram of the proposed system. The latter is composed
of two main parts: a cap made of flexible neoprene on which
custom made connectors stabilize the probes and a light control
module to be attached to the belt, that contains battery and
circuitry to acquire NIRS and EEG signals. The prototype
called “8-channel” gathers 8 scalp EEG electrodes recording
at 320 Hz sampling frequency, 8 light-emitting diodes as light
sources and 8 avalanche photodiodes (APD) as light detectors
for continuous-wave (CW) NIRS sensors recording at 20 Hz
for each channel. Every emitter can be coupled with 4 detectors
187
Fig. 2. Functional block diagram of the proposed EEG-NIRS system.
that lead to a total of 32 NIRS channels. The control module
has great portability through a small size, lightweight, and an
integrated Bluetooth device to wirelessly transmit in real time
acquired data to a graphical user interface (GUI).
The device can run an acquisition for at least 24 hours
without recharging the battery, thanks to low power dissi-
pation instrumentation like high-sensitivity APDs biased by
a high-voltage source of V, time-multiplexed light
emitting diode (LED) technique to minimize the illumination
duration and a low power FPGA platform. The 32 available
NIRS channels can image simultaneously one or more areas
of the cortex depending on the chosen protocol, increasing the
spatial resolution. Regarding the wireless link, it consists of
a Bluetooth Serial Port Profile (SPP) protocol able to provide
the required data throughput of 82 kbps with limited power
consumption. The proposed device meets the safety regulation
of IEC 60601-1.
The proposed wireless EEG-NIRS system includes a cap
equipped with sensor holders, sensors, control module and
graphical user interface. The architecture of the acquisition
chain is illustrated in Fig. 2 and details of each part of the
system are described below.
1) Cap and Sensors: The 8-channel prototype shown on
Fig. 3(a) contains 8 NIRS light emitters and 8 detectors. To
provide information on hemoglobin, small-size two-wavelength
LEDs (Epitex) emitting at 735 and 850 nm are used for their
low-power consumption compared to laser diodes. Each LED
is pulsed during a brief period (one thousandth of a second),
limiting the illuminated power. Avalanche photodiodes S2384
(Hamamatsu) are used as detectors to provide enough sensitivity
when biased by a high DC voltage, they are followed by a tran-
simpedance amplifier (TIA) circuit adjusted to deliver a gain of
10 MV/A with a 3 dB frequency cut-off of 84 kHz. The light
driver circuit and the photodiode and TIA circuit have been re-
duced on single round printed-circuit boards (PCBs) of 1.3 cm
[Fig. 3(b)–(d)] each.
These PCBs are packaged into custom-made 3D-printed
housings that are made to fit into connectors sewed on a
light-weighted dark elastic cap. The emitters and detectors
have the same size to improve modularity in the placement of
sensors. The neoprene-made cap is composed of 3-cm distant
188 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO. 2, APRIL 2013
Fig. 3. Illustration of the EEG-NIRS system. (a) Portable wireless device.
(b) Detector. (c) Source. (d) 3D views of the sensor where spring pushes the
sensor towards the scalp. (e) Portable control module. (f) Three layers of the
module. The battery is located in the power management board.
holes all around the surface to allow imaging different parts
of the cortex. Holes are useful for the clinician to remove
light-blocking hair before connecting the sensor. The optode
housing design contains a spring that pushes the sensor towards
the scalp to provide a good contact, limiting scattered lighting
through the cap (see Results section). Connectors and cap
designs have been optimized to reject ambient light, while
reducing discomfort. Scalp EEG electrodes can be put through
the cap and fixed using conductive paste. The installation of the
cap on a patient can take from 15 to 30 minutes depending on
their hair type, which affects the hair-clearing process. Patients
wore this cap for several hours without annoyance.
2) Control Module: The control module is divided into three
boards: the first one contains the acquisition circuits with analog
filtering, Analog-to-Digital Converters (ADCs) and Digital-to-
Analog Converters (DACs), a power management board and
a low-power commercially available FPGA board (low-power
reference platform from Arrow). The acquisition board acquires
samples from the amplified signals coming from the sensors on
the cap. An anti-aliasing filter precedes the NIRS signal digital-
ization at 20 Hz, while EEG is sampled at 320 Hz. Both modal-
ities are recorded with 16-bit resolution. The board owns also
the EEG amplification and reference circuits: the signal from
the EEG electrodes is amplified by 10000 before digitalization
through a 3-module chain, an instrumentation amplifier is used
at the front-end to minimize the noise, and two more amplifiers
filter the signal between 0.1 and 100 Hz. The EEG reference cir-
cuit requires two feedback signals from electrodes placed sym-
metrically on the mastoids to reduce the common mode noise,
and a reference electrode to place the head to a potential of
2.5 V. The second board provides 5 V analog and 3.3 V dig-
ital supplies from a 7.4 V 10 Ah Li-Ion battery (7.6 8 4 cm ,
400 g). A control circuit allows a high-voltage DC-DC regulator
(EMCO, CA02-5N) to provide a bias voltage from 0 to
V, to adjust the sensitivity of the avalanche photodiodes.
An Altera FPGA controls the converters through SPI buses so
that the duration of illumination is minimized, while meeting the
required bandwidth of the transimpedance amplifier. These du-
rations can be modified as they are sent by user interface before
an acquisition in a table containing the channels to be acquired,
the intensity of LEDs and timings. An embedded processor al-
lows user-controlled flexible configuration, acquisition of sam-
ples stored in a FIFO, data packaging, and communication on a
serial port.
The NIRS system has 32 channels and each channel is
composed of 2 wavelengths, thus 72 curves (32 NIRS times
2 wavelengths and 8 EEG curves) in real-time are transmitted
through a Bluetooth connection to a GUI developed on Lab-
view. The SPP protocol is provided by a commercially available
module (Parallax, Inc) to replace a serial cable. A custom pro-
tocol with header allows transmission error detection, which
can more often happen with a wireless link.
3) Graphical User Interface: In order to display in real-time
the 72 curves, a Labview-based software has been developed.
The interface allows the user to choose the channels to acquire,
and to select the photodiode bias voltage and the light intensity
of the emitters. After a first configuration, calibration tools help
to set the best parameters to obtain a good signal. These pa-
rameters can be modified during an acquisition. In order to ac-
quire the data from the prototype, a user’s interface calls upon
Boost libraries to implement a serial port and commu-
nicate through it. These libraries receive packages, fill curves
with data points and detect data loss, in which case interpola-
tion is made. It has been tested that an acquisition can run for
24 hours and the duration is only limited by the battery capacity.
Curves can be exported into Matlab format for off-time signal
processing.
B. Invasive Studies
The proposed invasive icEEG recording device is an im-
plantable wireless acquisition system used in patients with
drug-resistant focal epilepsy in which intracranial electrodes
are implanted to better delineate the epileptogenic zone. The
implantable device is targeted to be placed on top of the skull
SAWAN et al.: WIRELESS RECORDING SYSTEMS: FROM NONINVASIVE EEG-NIRS TO INVASIVE EEG DEVICES 189

aside from the craniotomy opening [Fig. 1(b)]. Through the

craniotomy window, many subdural grid, strip and depth elec-
trodes (diameter: 5 mm and inter-electrode contact spacing:
10 mm) are implanted to sample areas of suspected epilepto-
genicity. Teflon coated wires provide electrical connections
between many electrodes and the recording device, which is
mechanically tied to the skull with screws. No external wires
or exit wires are used through the skin, since a radio frequency
(RF) wireless link is established to the external equipments.
This method allows closing the craniotomy opening with the
patient’s skin and sealing bacteria entry ports.
This strategy addresses some of the drawbacks involved with
the traditional invasive epilepsy presurgical evaluation. The
wireless device reduces risks of infection and cerebrospinal
fluid leaks. Furthermore, removing wires between the external
base station and the electrodes improves the patient comfort
and mobility.
The wireless implantable device for icEEG recording re-
Fig. 4. Functional block diagrams of wireless icEEG system.
quires, during the presurgical evaluation, special needs and
reliable measurements equal to the traditional wired solution.
For the special needs, the acquisition circuitry is miniaturized
and moved from the external equipment to the implantable
circuits. The basic requirement to provide reliable intracra-
nial measurement is to have a low-noise and high-resolution
front-end preamplifier. The programmable low-noise amplifier
provides high gain while minimizing flicker and thermal noises
in the band of interest (0.1 to 2 kHz). The typical icEEG signal
amplitude varies from patient to patient (e.g., few V to mV)
and signal strength is dependent on electrode-tissues interface.
Following the preamplifier, an ADC is employed to convert
the icEEG signals to the digital domain using a low-noise Fig. 5. Photograph of the wireless icEEG recording device. (a) Implantable
internal reference voltage. The ADC should have sufficient part. (b) Remote controller.
resolution and sampling frequency ( kHz) to accommodate
icEEG measurements. The proposed system uses time-division
multiplexing capability for multiple channel recordings. Some take advantage of commercially available components suiting
sort of queuing is also necessary to permit framing of data the needs of raw icEEG wireless measurements.
and burst transmission wirelessly. Typically, transmitting data 1) Implantable Recording Device: The implantable
by burst instead of doing so continuously can help reduce recording device has front-end recording module and back-end
power consumption, and reduce interference susceptibility, wireless transceiver. Fig. 5(a) shows the implantable recording
while maximizing throughput. A certain amount of tunability device. The recording system is based on commercially avail-
is provisioned to permit parameter optimization and flexibility able microchip ADS1298 (Texas Instruments Incorporated)
toward different patient recordings. For this reason, a bi-di- that contains amplifiers, filters and digital converters. This
rectional wireless link is targeted, enabling remote control recording system features relatively high level of integration,
and tuning of the implantable device. A half-duplex link is tunability and low-power consumption. The amplifiers contain
implemented using a diplexer, a transmitter and a receiver. 8 differential programmable gain amplifiers providing high
An instruction encoding/decoding mechanism is employed for common-mode rejection ratio (CMRR) of 155 dB, while mini-
the downlink path, while the uplink path formats the data into mizing input-referred noise under 1 V RMS at the maximum
an optimized packet structure that is decoded at the remote gain setting. Additionally, 8 delta-sigma ADCs are available,
receiver. Remote instructions may be employed to modify offering 24 bits of resolution, up to 32 kHz sampling frequency
amplifier gain, channel selection, ADC sampling frequency per input, and 818 W power dissipation. The ADS1298
and resolution, and remotely control the acquisition. microchip accommodates low-power modes while consuming
The wireless icEEG system contains implantable recording 0.75 mW per channel. The proposed front-end system includes
device, remote controller and a GUI. Functional diagram of two ADS1298 chips, providing 16-channel recording per
the implantable device and corresponding remote controller are daughter board. Additional boards can be stacked to increase
all shown in Fig. 4. The device is implemented focusing on the number of analog recording channels.
building a solid framework, using commercially available de- The most critical part of this system is its wireless communi-
vices as building blocks, and designing a scalable system. We cation. A high-channel count EEG recording implant translates
190 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO. 2, APRIL 2013

into high-data rate and low-power specifications. Many com- TABLE I

mercial devices are available, but only few of them can meet MEASUREMENT FEATURES FOR THE WIRELESS NIRS-EEG SYSTEM
the high throughput and low-power specifications.
The back-end system contains Zarlink Medical Implant Com-
munication Service (MICS) transceiver. Zarlink’s ZL70102 and
ZL70321 (implant version of ZL70102) offer a complete radio-
telemetry solution in the 402–405 MHz MICS band at high-
data rate (800 kbps) and low-current consumption (5 mA). A
custom automatic-repeat request protocol, combined with ad-
vanced cyclic redundancy check (CRC), and forward-error cor-
rection (FEC), form a media access control (MAC) offering re-
liable communication . A 2.45 GHz receiver
is included to permit very low-power remote wakeup from the
base station. A printed loop antenna is also used, offering a
dBi gain in the 400 MHz band. The ZL70321 device is dig-
itally controlled, through a SPI interface, with a TI MSP430 Mi-
crocontroller, and permits integration with the analog front-end.
The proposed implantable part contains a custom board inte-
grating the RF circuits with the host microcontroller, other dig-
ital components and a rechargeable 1200 mAh Li-ion battery
(35 mm 18 mm, 35 g) for limited channel transmission or
2800 mAh Li-ion battery (56 mm 18 mm, 50 g) for maximum
data transmission.
2) Remote Controller: The remote-controller board uses
the Zarlink ZL70102 transceiver to permit remote wakeup and
control of the implant. The remote-control board in Fig. 5(b)
contains a 402–405 MHz MICS transceiver and 2.45 GHz
transmitter controlled by a TI MSP430 Microcontroller. The
controller board also provisions a USB data link with the base
station computer to permit recording the raw EEG waveforms
Fig. 6. Validation of wireless EEG-NIRS system. (a) Test subject wearing the
transmitted by the implant, and to forward remote commands device. (b) Visual event-related potentials protocol for the test subject.
from base station to the implant.
3) Graphical User Interface: One of the key aspects of the
system is tuneability and control for the icEEG acquisition, 40 mm PCB and remote controller in Fig. 5(b). Validation of
through the use of a mostly digital architecture. Therefore, the EEG-NIRS and icEEG system are demonstrated below.
it involves an extensive use of software at various levels, to
partition the workload between the different nodes. A GUI is A. Validation of the Wireless NIRS-EEG System
developed using C# program and. NET development tools for Several in-vivo protocols have been established to validate
easy control, fine tuning and multichannel icEEG recordings. the NIRS signals; Table I illustrates measured features. Fig. 6(a)
During the icEEG acquisition, this system needs to handle shows a test subject wearing flexible cap with sensors and the
many sets of data periodically and quickly. For this reason, a control module attached to his waist belt without reducing com-
high-priority worker thread handles receiving and verifying the fort. Different subjects participated in the device’s validation.
data to ensure real-time operation. It also writes samples to a Fig. 6(b) shows the performed visual test, which is described
local file and update progress on the GUI. This architecture later.
permits fast handling of the incoming data on the base station In order to validate the results, we compared the mea-
while maintaining a responsive GUI to the end-user. surements of oxyhemoglobin (HbO ) and de-oxyhemoglobin
(HbR) concentrations on the arm during a constriction using an
III. RESULTS inflatable cuff from a sphygmo-manometer with a commercially
The noninvasive 8-channel wireless EEG-NIRS device is available device (ISS Imagent). Hemoglobin concentrations
shown in Fig. 3(a). Dimensions of the control module are have been calculated with the Matlab toolbox HomER [8].
16 13 8.2 cm and weight is 800 g; the box can be at- Figs. 7(a) and (b) present the results from both devices.
tached to the human body to reduce discomfort. Cap is made Both curves show stable concentrations before the constric-
of neoprene. The diameters of each detector [Fig. 3(b)] and tion (region 1), then an increase in blood flow. At the end of
illuminator [Fig. 3(c)] are 11 mm. Fig. 3(d) shows 3D model of the inflation (region 2), HbO decreases while HbR increases
the sensors encapsulation. The control module and Bluetooth due to the non-renewal of the blood. Finally, when the pressure
device are depicted in Fig. 3(e) and 3D model of the module are is released (region 3), the blood starts flowing again and both
shown in Fig. 3(f). On the other hand, Fig. 5(a) shows the inva- concentrations stabilize (region 4). The difference between the
sive wireless icEEG system that was implemented on 55 mm curves is due to the duration of the constriction, which has been
SAWAN et al.: WIRELESS RECORDING SYSTEMS: FROM NONINVASIVE EEG-NIRS TO INVASIVE EEG DEVICES
Fig. 7. Averaged relative changes of concentration (dConcentration) of HbO
(continuous line) and HbR (dotted line) using (a) commercially available device,
and (b) the proposed wireless NIRS system. Y-axis are qualitative evaluations,
tests have not been done at the same time or with the same optode distance.
longer for the commercial prototype. It explains why the con-
centrations diverge more. Thus, the commercially available de-
vice and the proposed prototype show similar results after cal-
culation of hemoglobin concentrations.
Several tests consisting of imaging the cerebral cortex have
been useful to validate the NIRS signals. Fig. 8(a) presents
a typical NIRS signal on which clearly appear the heartbeat
and the Mayer wave. Two types of protocols have been estab-
lished: stimulation of visual cortex using a flashing pattern,
and of motor cortex with finger tapping. The event-related
tasks consist of 30 sec of baseline, then 10 iterations of 30 sec
stimulation followed by a 30 sec rest period. Fig. 8(b) shows
the sensor placement and the average of the evolution of
HbO concentrations during the visual task. As expected, the
concentration increases at the beginning, reaches peak ampli-
tude, plateaus, then decreases towards baseline as the task ends.
Fig. 8(c) shows concentrations in HbO and HbR recorded
curves for visual tasks revealing an increase of HbO and a
decrease of HbR during stimulation in the occipital region,
as expected. Mapping of 2-dimension reconstructed activity
is shown in Fig. 8(d). Placement of sensors for motor task is
presented in Fig. 8(e), and motor cortex activation curves in
Fig. 8(f). Table II shows a comparison between the wireless
191
NIRS-EEG system and other relevant works in the literature.
The proposed system has maximum numbers of channels
in NIRS and EEG system. Also, it has moderate recording
sampling rate for the application and a 16-bit resolution.
B. Validation of Wireless IcEEG System
Measurements on each part of the wireless icEEG system
were done separately. Fig. 9(a) shows measured input-referred
noise voltage while recording at 250 Hz sampling frequency and
the noise voltage is 0.5 Vrms. The measured input-referred
noise voltage spectral density of the bioamplifier is shown in
Fig. 9(b).
The corner frequency is located around 0.6 Hz and the
thermal noise level equals to approximately 0.5 nV Hz .
Integration under this curve gives 0.5 Vrms noise voltage,
which is consistent with measured noise voltage. Table III
shows further measurements for the wireless icEEG system.
Following the measurements, the wireless intracranial
system was validated using archived icEEG recordings from
two patients with intractable non-lesional partial epilepsy
who had previously undergone an intracranial study to better
delineate their epileptogenic zone. This study was conducted
at Notre-Dame Hospital, Centre Hospitalier de l’Université
de Montréal (CHUM) with approval from the CHUM ethical
committee. Initial icEEG recordings were performed using
commercially available amplifiers (PRO-36 amplifiers, Stellate
Harmonie System) during the 3 weeks epilepsy presurgical
evaluation. Offline 16-channel recordings from suspected areas
of epileptogenicity were uploaded onto a signal conditioner
(Arbitrary signal waveform generator) to reproduce icEEG
data and fed into the wireless icEEG system for continuous
wireless icEEG transmission to the remote computer. The
wireless icEEG system stopped recording when system moved
outside of the transmission range and received data started
deteriorating. The wirelessly transmitted icEEG recording
quality was evaluated in terms of distortion between the orig-
inal and recorded waveforms; the normalized root-mean-square
deviation (NRMSD) was calculated. The average NRMSD is
2% during the continuous multiple channels wireless icEEG
recording. Lifetime of the invasive EEG monitoring system
is approximately 3 weeks and a Lithium battery supports this
recordings. There is no recharging option available for the
presented system.
The original intracranial recordings from both patients con-
tained several seizures, originating from the right hippocampus
and the lateral temporal neocortex respectively based on visual
analysis from an expert epileptologist. The icEEG data which
were fed into the wireless icEEG were analyzed using a seizure
detection system previously described in [5]. Fig. 10(a) illus-
trates the wireless icEEG recording and monitoring procedure
on the first patient (24 years old male).
Among the 16 channels, Ch#1–Ch#4 electrodes were located
in the lateral orbitofrontal region (grid electrodes), Ch#5 and
Ch#6 in the right hippocampus (depth electrodes), Ch#7–Ch#8
in the frontal lobe (strip electrodes), and Ch#9–Ch#16 in the
right insula (depth electrodes). The automated seizure detection
algorithm first detected seizure activity from Ch#5 and Ch#6
192 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO. 2, APRIL 2013

Fig. 8. Measured NIRS recordings. (a) NIRS baseline signal. (b) Sensor placement and average HbO concentrations for the visual task. (c) Averaged relative
changes of concentration (dConcentration) of HbO and HbR for the visual task. (d) Topographic reconstruction of hemodynamic changes. (e) Placement of sensors
for motor task. (f) Averaged relative changes of concentration (dConcentration) of HbO and HbR for motor task.

TABLE II TABLE III

A COMPETITIVE COMPARISON OF DIFFERENT TYPES OF MEASUREMENT FEATURES FOR THE WIRELESS ICEEG SYSTEM
NIRS-EEG PORTABLE SYSTEM

right insula [Fig. 10(b)]. The icEEG recordings of this patient

Fig. 9. Measured noise at the front-end of icEEG recording device. (a) Input-re-
ferred noise voltage. (b) Measured input-referred noise voltage spectral densi-
ties.
over the right medial temporal lobe (hippocampus) with subse-
quent spreading to the right lateral temporal neocortex and the
revealed that all eight recorded seizures originated from Ch#5
and Ch#6.
The reasonable assumption was concordant with the clin-
ician’s visual analysis. Similarly for the second patient
(36-year-old male), the automated seizure detection system
adequately identified the seizure onset zone over the lateral
temporal neocortex using the output icEEG data when fed
through the wireless system. Table IV shows a comparison
between the wireless icEEG system and other relevant works
SAWAN et al.: WIRELESS RECORDING SYSTEMS: FROM NONINVASIVE EEG-NIRS TO INVASIVE EEG DEVICES 193

Furthermore, this work has lowest BER compared

to other types of wireless technology, thus it provides highest
accuracy and reliability for wireless icEEG recordings in order
to indentify epileptogenic zone prior to a brain surgery.

IV. CONCLUSION
We have demonstrated a wireless portable EEG-NIRS
system for functional brain imaging that allows the trans-
mission of 8 EEG and 32 NIRS channels through Bluetooth
connection. The preliminary experimental results have showed
enough sensibility to observe hemodynamic activations and
adequate autonomy for 24-hour. Also, we have presented a new
multichannel wireless icEEG recording system for epilepsy
presurgical evaluation. We demonstrated that proposed im-
plantable module offers great tuneability, with programmable
gain, input-referred noise of 0.5 VRMS, sampling frequency
up to 2 ks/s for 32 channels of uncompressed and low dis-
tortion (NRMSD under 2%) icEEG signals. The C# program
based GUI offers easy remote control and tuneability over the
implantable system, communicating in both directions with a
MICS band transceiver. The RF link has shown suited relia-
bility with corrected BER lower than for a raw channel
quality of . We also have demonstrated that proposed
architecture is scalable, and the 128 channels target could be
obtained with a custom System-on-chip device.

Fig. 10. The icEEG monitoring procedure. (a) icEEG recorded from different
locations in patients. (b) Early seizure onset detection allows localizing the
epileptogenic zone.
TABLE IV
A COMPETITIVE COMPARISON OF DIFFERENT TYPES OF WIRELESS SYSTEM
in the literature. This work supports both MICS and ISM
bands; however, others work in Table IV support single carrier
frequency. Therefore, the presented wireless device will open
a new range of clinical applications for the next generation of
medical implantable devices. Moreover, this wireless icEEG
recording device features relatively moderate data rate and
transmission range with lower power dissipation, which is
suitable for the epilepsy presurgical evaluation due to the
limitation in craniotomy area and implanted electrodes count.
REFERENCES
[1] M. Fifer, S. Acharya, H. Benz, M. Mollazadeh, N. Crone, and N.
Thakor, “Toward electrocorticographic control of a dexterous upper
limb prosthesis: Building brain-machine interfaces,” IEEE Pulse, vol.
3, no. 1, pp. 38–42, Jan. 2012.
[2] I. H. Stevenson and K. P. Kording, “How advances in neural recording
affect data analysis,” Nature Neurosci., vol. 14, pp. 139–142, 2011.
[3] R. F. Yazicioglu, T. Torfs, P. Merken, J. Penders, V. Leonov, R. Puers,
B. Gyselinckx, and C. VanHoof, “Ultra-low-power biopotential inter-
faces and their applications in wearable and implantable systems,” Mi-
croelectron. J., vol. l40, pp. 1313–1321, 2009.
[4] S. S. Spencer, D. K. Nguyen, and R. B. Duckrow, “Invasive EEG in
presurgical evaluation of epilepsy,” in The Treatment of Epilepsy, S.
Shorvon, E. Perucca, and J. Engel, Eds., 3rd ed. Hoboken, NJ, USA:
Wiley–Blackwell, 2009, pp. 767–798.
[5] M. T. Salam, D. K. Nguyen, and M. Sawan, “A multichannel intrac-
erebral EEG monitoring system for epilepsy presurgical evaluation,”
in Proc. IEEE Canadian Conf. Electrical and Computer Engineering,
Niagara, ON, Canada, 2011.
[6] M. T. Salam, M. Sawan, and D. K. Nguyen, “Low-power implantable
device for onset detection and subsequent treatment of epileptic
seizures: A review,” J. Healthcare Eng., vol. 1, no. 2, pp. 169–184,
2010.
[7] A. Villringer and B. Chance, “Non-invasive optical spectroscopy and
imaging of human brain function,” Trends Neurosci., vol. 20, pp.
435–42, 1997.
[8] T. J. Huppert, S. G. Diamond, M. A. Franceschini, and D. A. Boas,
“HomER: A review of time-series analysis methods for near-infrared
spectroscopy of the brain,” Appl. Opt., vol. 48, no. 10, pp. 280–298,
2009.
[9] D. K. Nguyen, J. Tremblay, P. Pouliot, P. Vannasing, O. Florea, L.
Carmant, F. Lepore, M. Sawan, F. Lesage, and M. Lassonde, “Non-
invasive continuous EEG-fNIRS recording of temporal lobe seizures,”
Epilepsy Res., vol. 99, no. 1–2, pp. 112–26, 2011.
[10] J. Furusho, A. Suzuki, Y. Takakusa, F. Kawaguchi, N. Ichikawa, and
T. Kato, “Simultaneous study of interictal EEG and near-infrared spec-
troscopy in a boy with epilepsy,” Int. Congress Ser., vol. 1232, pp.
673–676, 2002.
[11] G. Strangman, R. Goldstein, S. L. Rauch, and J. Stein, “Near-infrared
spectroscopy and imaging for investigating stroke rehabilitation: Test-
retest reliability and review of the literature,” Arch. Phys. Med. Re-
habil., vol. 87 (Suppl. 2), pp. 12–19, 2006.
194 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO. 2, APRIL 2013
[12] E. Lareau, F. Lesage, P. Pouliot, D. Nguyen, J. Le Lan, and M. Sawan,
“Multichannel wearable system dedicated for simultaneous electroen-
cephalography/near-infrared spectroscopy real-time data acquisi-
tions,” J. Biomed. Opt., vol. 16, no. 9, 2011, doi:10.1117/1.3625575.
[13] M. Sawan, M. T. Salam, S. Gelinas, J. Le Lan, F. Lesage, and D. K.
Nguyen, “Combined NIRS-EEG remote recordings for epilepsy and
stroke real-time monitoring,” in Proc. IEEE Int. Symp. Circuits and
Systems, Seoul, Korea, May 2012.
[14] R. Harrison, P. Watkins, R. Kier, R. Lovejoy, D. Black, B. Greger,
and F. Solzbacher, “A low-power integrated circuit for a wireless 100-
electrode neural recording system,” IEEE J. Solid-State Circuits, vol.
42, no. 1, pp. 123–33, Jan. 2007.
[15] M. Rizk, C. A. Bossetti, T. A. Jochum, S. H. Callender, M. A.
L. Nicolelis, D. A. Turner, and P. D. Wolf, “A fully implantable
96-channel neural data acquisition system.,” J. Neural Eng., vol. 6,
no. 2, p. 026002, Apr. 2009.
[16] B. Gosselin, A. Ayoub, J.-F. Roy, M. Sawan, F. Lepore, A. Chaudhuri,
and D. Guitton, “A mixed-signal multichip neural recording interface
with bandwidth reduction,” IEEE Trans. Biomed. Circuits Syst., vol. 3,
no. 3, pp. 129–141, Jun. 2009.
[17] H. Miranda, V. Gilja, C. Chestek, K. Shenoy, and T. Meng, “Hermesd:
A high-rate long-range wireless transmission system for simultaneous
multichannel neural recording applications,” IEEE Trans. Biomed. Cir-
cuits Syst., vol. 4, no. 3, pp. 181–191, Jun. 2010.
[18] R. Harrison, R. Kier, C. Chestek, V. Gilja, P. Nuyujukian, S. Ryu, B.
Greger, F. Solzbacher, and K. Shenoy, “Wireless neural recording with
single low-power integrated circuit,” IEEE Trans. Neural Syst. Rehabil.
Eng., vol. 17, no. 4, pp. 322–329, Aug. 2009.
[19] M. T. Salam, M. Sawan, and D. K. Nguyen, “A novel low-power im-
plantable epileptic seizure-onset detector,” IEEE Trans. Biomed. Cir-
cuits Syst., vol. 5, no. 6, pp. 568–578, Oct. 2011.
[20] H. Kida, M. Fujii, T. Inoue, Y. He, Y. Maruta, S. Nomura, K.
Taniguchi, T. Ichikawa, T. Saito, T. Yamakawa, and M. Suzuki,
“Focal brain cooling terminates the faster frequency components of
epileptic discharges induced by penicillin G in anesthetized rats.,”
Clin. Neurophysiol., vol. 123, no. 9, pp. 1708–13, 2012.
[21] M. T. Salam, M. Mirzaei, M. Sawan, and D. K. Nguyen, “A im-
plantable asynchronous responsive drug delivery system for treatment
of epilepsy,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 20, no. 4,
pp. 432–442, 2012.
[22] M. J. Morrell, “Responsive cortical stimulation for the treatment of
medically intractable partial epilepsy,” Neurology, vol. 77, no. 13, pp.
1295–304, 2011.
[23] I. Osorio and M. G. Frei, “Real-time detection, quantification, warning,
and control of epileptic seizures: The foundations for a scientific epilep-
tology,” Epilepsy Behav., vol. 16, pp. 391–396, 2009.
[24] M. Stead, M. Bower, B. H. Brinkmann, K. Lee, W. R. Marsh, F. B.
Meyer, B. Litt, J. V. Gompel, and G. A. Worrell, “Microseizures and
the spatiotemporal scales of human partial epilepsy,” Brain, vol. 133,
pp. 2789–2797, 2010.
[25] E. Chua, W. C. Fang, C. K. Chen, C. C. Fu, S. Y. Tseng, S. Kang, and
Z. H. Fsieh, “A highly-integrated biomedical multiprocessor system
for portable brain-heart monitoring.,” in Proc. IEEE Int. Symp. Circuits
and Systems, 2011, pp. 1532–1535.
[26] L. Rovati, G. Salvatori, L. Bulf, and S. Fonda, “Optical and electrical
recording of neural activity evoked by graded contrast visual stimulus,”
Biomed. Eng. Online, vol. 6, no. 28, 2007.
[27] C. K. Liang, J. J. Chen, C. L. Chung, C. L. Cheng, and C. C. Wang,
“An implantable bi-directional wireless transmission system for tran-
scutaneous biological signal recording,” Phys. Meas., vol. 26, no. 1,
pp. 83–97, 2005.
[28] S. Lee, J. Yoo, S. Kiseok, and Y. Hoi-Jun, “A 1.3 pJ/b inductive cou-
pling transceiver with adaptive gain control for Cm-range 50 Mbps data
communication,” in Proc. IEEE Asian Solid-State Circuits Conf., 2009.
[29] Texas Instrumentations: CC2500, Feb. 20, 2013 [Online]. Available:
http://www.ti.com/lit/ds/symlink/cc2500.pdf
[30] S. Farshchi, P. H. Nuyujukian, A. Pesterev, I. Mody, and J. W. Judy,
“A tinyos-enabled mica2-based wireless neural interface,” IEEE Trans.
Biomedical Eng., vol. 53, no. 7, pp. 1416–1424, 2006.
[31] R. Fazel-Rezai, M. Pauls, and D. Slawinski, “A low-cost biomedical
signal transceiver based on a bluetooth wireless system.,” in Proc.
IEEE Eng Med. Biol. Soc. Conf., 2007, vol. 2007, pp. 5712–5715
[Online]. Available: http://dx.doi.org/10.1109/IEMBS.2007.4353643
[32] A. P. Chandrakasan, F. S. Lee, D. D. Wentzloff, V. Sze, B. P. Ginsburg,
P. P. Mercier, D. C. Daly, and R. Blazquez, “Low-power impulse uwb
architectures and circuits,” Proc. IEEE, vol. 97, no. 2, pp. 332–352,
2009.
[33] Alereon: AL5100/AL5301 Worldwide Ultrawideband Chipset,
Feb. 20, 2013 [Online]. Available: http://www.alereon.com/prod-
ucts/chipsets
Mohamad Sawan (S’88–M’89–SM’96–F’04)
received the Ph.D. degree in electrical engineering
from Sherbrooke University, Sherbrooke, QC,
Canada, in 1990.
He joined the École Polytechnique de Montréal,
Montreal, QC, Canada, in 1991, where he is currently
a Professor of microelectronics and biomedical en-
gineering. He is a Canada Research Chair in Smart
Medical Devices, and he is leading the Microsystems
Strategic Alliance of Quebec (ReSMiQ). He is the
Founder of the Polystim Neurotechnologies Labora-
tory. He has authored or co-authored more than 600 peer-reviewed papers, two
books, 10 book chapters, and holds 12 patents. His current interests include de-
sign and test of analog, digital, RF, microelectromechanical systems and optic
circuits, and microsystems.
Dr. Sawan is Deputy Editor-in Chief of the IEEE TRANSACTIONS ON
CIRCUITS AND SYSTEMS II—EXPRESS BRIEFS and Associate Editor of the IEEE
TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS. He has received
several awards, among them the Bombardier Award for technology transfer,
and the Barbara Turnbull Award for spinal-cord research. He is a Fellow of the
Canadian Academy of Engineering, a Fellow of the Engineering Institute of
Canada, and an Officer of the Quebec’s National Order.
Muhammad T. Salam received the B.A.Sc. degree
in electrical and electronics engineering from the
Islamic University of Technology, Board Bazar,
Bangladesh, and the M.A.Sc. degree in electrical and
computer engineering from Concordia University,
Montreal, QC, Canada, in 2003 and 2007, respec-
tively.
Currently, he is working toward the Ph.D. degree
in electrical engineering at École Polytechnique de
Montréal, Montreal, QC, Canada. He is employed
in the Polystim Neurotechnologies Laboratory and
Epilepsy Monitoring Unit, Notre-Dame Hospital, Montreal, QC, Canada,
where his research focuses on implantable micro-device.
Jérôme Le Lan graduated in electrical engineering
from École Supérieure d’eÉlectricité—Supélec,
Paris, France, in 2011 after having studied advanced
mathematics–physics in “classes préparatoires.”
He joined Polystim Neurotechnologies Labora-
tory, Montreal, QC, Canada, to complete the M.A.Sc.
degree in biomedical engineering, by working on the
development of a 32-channel NIRS-EEG portable
prototype for brain imaging.
Amal Kassab received the B.A.Sc. degree in me-
chanical engineering from Concordia University,
Montreal, QC, Canada, in 2011.
Currently, she is working toward the M.A.Sc.
degree in biomedical engineering at École Poly-
technique de Montréal, Montreal, QC, Canada. Her
work with Polystim Neurotechnologies, Montreal,
QC, Canada, focuses on the design of a combined
NIRS/EEG cap for long term, wireless brain activity
monitoring needed to examine epilepsy patients.
SAWAN et al.: WIRELESS RECORDING SYSTEMS: FROM NONINVASIVE EEG-NIRS TO INVASIVE EEG DEVICES
Sébastien Gélinas received the B.A.Sc. degree
in electrical engineering from Université Laval,
Quebec City, QC, Canada, in 2006.
Currently, he is working toward the M.A.Sc. de-
gree in electrical engineering at École Polytechnique
de Montréal, Montreal, QC, Canada. His work with
Polystim Neurotechnologies, Montreal, QC, Canada,
focuses on low-power wireless systems for brain
imaging applications. He interned at École Nationale
Supérieure des Télécommunications, Paris, France,
in 2005 and at Université Laval in 2002, focusing on
image and signal processing techniques.
Phetsamone Vannasing received the Electrophysi-
ology degree (E.P.M) from Ahuntsic College, Mon-
treal, QC, Canada, in 1989.
After graduation, she worked in different electro-
physiology fields with pediatric and adult popula-
tions. She joined CHU Sainte-Justine mère-enfant,
Montreal, QC, Canada, in 1993. She has worked
as a Research Coordinator at the optical imaging
laboratory as well as the electrophysiology sensory
and cognitive laboratories. Her work consists of
teaching technical aspects and data analyses of NIRS
and electrophysiology to doctoral and master’s degree candidates.
Frédéric Lesage received the Ph.D. degree in
theoretical physics at Université Paris-Sud, Orsay,
France, and CEA Saclay, Gif-sur-Yvette, France.
Currently, he is an Associate Professor in the
Department of Electrical Engineering at École
Polytechnique de Montréal, Montreal, QC, Canada.
He did a postdoctoral fellowship at the University
of Southern California, Los Angeles, CA, USA. He
then worked for Advanced Research Technologies
Inc. in molecular imaging. His research interests
include molecular imaging based on fluorescence
and photoacoustic, and brain imaging using hemodynamic signals and diverse
195
techniques, namely: diffuse optical imaging, OCT, two-photon microscopy and
intrinsic optical imaging.
Maryse Lassonde received the B.S. degree in psy-
chology from the Université de Montréal, Montreal,
QC, Canada, and the Ph.D. degree from Stanford
University, Stanford, CA, USA.
She became a Professor at the Université du
Québec à Trois-Rivières (UQTR, 1977–1988),
Trois-Rivières, QC, Canada, and then at the Uni-
versité de Montréal. Since the beginning of her
career, she has been associated with several research
centers, among them the Sainte-Justine University
Hospital Center and the Centre de Recherche en
Neuropsychologie et Cognition. She has authored five books, more than 200
book chapters, and articles in scientific journals.
Dr. Lassonde is a Fellow of the Canadian Psychological Association (1994),
of the Royal Society of Canada (1997), of the Canadian Academy of Health
Sciences (2010), and holds a Canada Research Chair in Developmental Neu-
ropsychology since 2001. She has received several other awards and distinc-
tions. Among them, she was named Knight, National Order of Quebec in 1999
and Officer of the Order of Canada in 2012. She is a member of several editorial
committees and was named honorary professor at the University of Auckland
and at the Université de Paris V in 2007. In January 2012, she became the Sci-
entific Director of the Quebec Science and Technology granting agency.
Dang K. Nguyen received the M.D. degree from the
Université de Montréal, Montreal, QC, Canada.
Currently, he is an Associate Professor of medicine
at the Université de Montréal, where he completed
his neurology residency, with expertise in epilepsy.
He practices at Notre-Dame Hospital, Montreal, QC,
Canada, where he acts as the Director of the Epilepsy
Monitoring Unit. His research interests focus on the
study of medically intractable epilepsies. He and
collaborators are developing and evaluating novel
methods to better localize the epileptogenic zone
and allowing its surgical resection in refractory cases.