28 Journal of Rehabilitation Robotics, 2013, 1, 28-41

Progress and Prospects in EEG-Based Brain-Computer Interface:

Clinical Applications in Neurorehabilitation

Sergio Machado1,2,3,4,*, Leonardo Ferreira Almada3 and Ramesh Naidu Annavarapu5

1
Laboratory of Panic and Respiration, Institute of Psychiatry of Federal University of Rio de Janeiro
(IPUB/UFRJ), Brazil; National Institute for Translational Medicine (INCT-TM, CNPq), Rio de Janeiro, Brazil
2
Physical Activity Neuroscience Laboratory, Physical Activity Sciences Postgraduate Program of Salgado de
Oliveira University (UNIVERSO), Niterói, Brazil
3
Institute of Phylosophy (IFILO), Federal University of Uberlândia (UFU), Minas Gerais, Brazil
4
Quiropraxia Program of Faculty of Medical Sciences, Central University (UCEN), Santiago, Chile
5
Department of Physics, School of Physical, Chemical and Applied Sciences, Pondicherry University,
Puducherry – 605 014, India
Abstract: Several patients are no longer able to communicate effectively or even interact with the outside world in ways
that most of us do it. For instance, severe cases astetraplegic or post-stroke patients are literally 'locked in' their bodies,
unable to exert any motor control after a spinal cord injury or a brainstem stroke, requiring alternative methods of
communication and control. However, in the near future, their brains may offer them a way out. EEG-based brain-
computer interface (BCI) is the technique utilized to measure brain activity and by the way that different brain signals are
translated into commands that control an effector (e.g., controlling a spelling system via eye movements). Here,we aim
to review the basic concepts of EEG-based BCI and the main advances in communication, in motor control restoration
and in downregulation of cortical activitythat seem to be relevant for clinical applications in the coming years
forneurorehabilitation of severely limited patients. It allows brain-derived communication in patients with amyotrophic
lateral sclerosis and motor control restoration in patients after spinal cord injury and stroke. In addition, epilepsy and
attention deficit and hyperactive disorder patients were able to downregulate their cortical activity. Owing to the rapid
progression of EEG-based BCI research over the last few years and the swift ascent of computer processing speeds and
signal analysis techniques, we suggest that emerging ideas related to clinical neurorehabilitation of severely limited
patients will generate viable clinical applications in the near future.

Keywords: Amyotrophic Lateral Sclerosis, attention deficit and hyperactive disorder, EEG-based brain-computer
interface, epilepsy, neurorehabilitation, spinal cord, stroke.

INTRODUCTION computers, switches or prostheses [4, 6-9]. The

The natural form of any communication or control
requires the proper functioning of peripheral nerves
and muscles.Any damage to them results in severe
motor disabilities and thus requires alternative means
of communication and control. Because, patients with
damage to their nerves and muscles are unable to use
effective means of communication which require
voluntary muscular control. Several studies over the
past two decades indicate that scalp-recorded
electroencephalogram (EEG) activity can be the basis
for such non-muscular communication and control
systems, commonly called brain-computer interfaces
(BCIs) [1-5].
EEG-based BCI is a communication system that
extracts specific features online and automatically from
EEG signals that can be detected on the scalp, and
uses these to operate external devices, such as
*Address correspondence to this author at the Rua Bolivar, 150/apt. 702,
Copacabana, CEP: 22061-020, Rio de Janeiro – RJ, Brazil;
Tel: +5521 91567006; E-mail: secm80@yahoo.com.br
messages or commands that a user wishes to convey
pass not through the brain’s normal output pathways,
but are instead extracted directly from brain signals [2,
10]. The basis for this phenomenon is that mental
activity (e.g. thought) is directly reflected in bioelectrical
brain activity and is therefore encoded in recorded
signals [1, 5, 11].
A BCI uses features in these signals to allow a
subject to communicate with the outside world [5, 9,
12]. The most commonly used features have been
sensorimotor rhythms (SMR) [13, 14], slow cortical
potentials (SCP) [6, 15-17] and the P300 event-related
potential (ERP) [13, 18]. Systems based on SMRs or
SCPs use components in the frequency or time domain
that are spontaneous in the sense that they are not
dependent on specific sensory events. On the other
hand, systems based on the P300 response use time-
domain EEG components that are elicited by specific
stimuli [5, 10, 13, 16]. The target populations for clinical
applications of BCI consist mainly of patients with
amyotrophic lateral sclerosis (ALS), severe CNS

E-ISSN: 2308-8354/13 © 2013 Synergy Publishers

Progress and Prospects in EEG-Based Brain-Computer Interface
damage, such as spinal cord injuries and stroke, and
epilepsy and attention deficit and hyperactive disorder
(ADHD), resulting in substantial deficits in
communication and motor function [7, 13, 19].
Thus, the potential utility of BCI for severely limited
patients’ applications will be to provide alternative and
supplemental control to communication, motor control
restoration, and down-regulation of cortical activity.
With this in mind, the present paper reviews both basic
concepts (main elements and types of BCI) and
presents the main advances in communication, motor
control restoration and down-regulation of cortical
activity, which appear to be relevant for clinical
applications in the coming years, particularly in the
context of rehabilitation of severely limited patients.
MAIN COMPONENTS OF BCI
BCI can measure the electrical activity of the brain,
process it, and produce control signals that reflect the
user’s intention. Several elements must be considered
when designing a BCI system (Figure 1). The first step
is designing an appropriate experimental protocol to
suit the application and the environment in which the
BCI will be used. This protocol includes a choice of
mental task, stimulus parameters (e.g., visual stimulus
timing and constraints), and minimization of unwanted
Figure 1: Basic design of BCI.
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 29
stimuli and distractions that may affect the properties of
the signals to be monitored.
The second step is establishing the type of
feedback to be sent to the user, although this step is
not yet done explicitly in many BCIs. Further, although
we do know that the user adapts to the entire system
and that the device’s response plays a role in this
process, user adaptation mechanisms in the context of
BCIs have only recently become a topic of investigation
[20-25]. These studies have shown that not only good
performance can lead to an achievement of BCI
system control but also a poor performance can lead to
achieved control, by co-adaptation of the user and the
system [25].
Finally, the last step is choosing the best EEG-
based BCI, which will be chosen based on how the
brain activity is recorded, how subjects are trained, how
the signals are translated into device commands, and
which application is provided to the users.
Mental Strategy
To control a BCI, a subject needs to engage in
mental activity. This mental activity must reliably be
reflected in measured EEG signals on the scalp.
Moreover, because a sophisticated brain-based control
30 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
system requires multiple possible control signals, a
subject must be able to engage in multiple mental
activities that can be distinguished quickly on the basis
of corresponding brain activities [26]. But not all mental
activity can give rise to such sustained, stable, and
controllable brain signals. Certain strategies for mental
processing work better than others, and work has been
done to evaluate which mental tasks can be used in
BCIs [27].
The most widely used mental tasks are motor
imagery (MI) [28] focused visual attention [29] and
operant conditioning [6, 16]. A widespread approach
uses MI of different parts of the body [30, 31]. Because
of the somatotopic organization of the primary motor
cortex, imagined movements cause characteristic
spatial activation patterns that are distinguishable by
classification algorithms and, therefore, can be used to
control an external effector by voluntary modulation of
the amplitude of different frequency bands, mainly mu
or beta bands [14, 32].
Another group of approaches uses focused visual
attention (self-regulation of brain-activity), which can be
learned by neurofeedback training. The voluntary up-
and down-regulation of slow-cortical potentials [33] via
an arbitrary subject-dependent strategy provides
appropriate control signals. Distinct from the
aforementioned self-paced approaches are those using
brain activity elicited as a response to external stimuli:
among these externally-paced BCIs are those based
on the steady state visual evoked potential (SSVEP
allows the subject to select one out of several
simultaneously presented flickering visual stimuli) [34,
35] and the P300 (elicited using odd-ball paradigms)
[17].
Output and Feedback Methods
Whichever mental strategies and brain activity
markers are used, subjects generally need training
before they have sufficient control over their brain
activity. Training involves not only the simple practice
of the mental tasks used but also an implicit fine-tuning
of brain activity that can only be established by implicit
learning using real-time feedback of ongoing brain
activity. Feedback is thus a very important component
in the training phase, as well as during application.
Most BCIs to date present their output (feedback) on a
computer screen in the form of letters, icons, or arrows.
This facilitates training, but the output from a BCI has
no clear restrictions and currently includes robotic arm
movements, wheelchair control. Some studies have
Machado et al.
begun to investigate the possibility of controlling
neuroprostheses via a BCI [5, 29, 36-38]. Feedback, or
output, can be discrete or continuous, realistic (e.g.
hand grasp), or virtual reality [9, 38].
In two experiments, attempts were made to control
either a real hand (via neuromuscular electrical
stimulation) or an artificial hand in users with a
complete lesion of the cervical spinal cord [39, 40] with
some success. However, some researchers have
pointed out that BCIs are not yet reliable or fast enough
to be used with some neural prostheses. For example,
BCIs are not yet useful for controlling walking induced
by functional neuromuscular electrostimulation (FES)
due to their current low reliability, which has not
improved much since a 2003 report by Sinkjaer et al.
[41]. Indeed, were a BCI used for controlling gait, at the
time of this writing, the user would risk falling every few
steps. Although recent studies have begun to evaluate
the use of vibrotactile feedback [34, 42], most studies
to date have employed only visual or auditory feedback
presentations. But within the visual and auditory
domains, different types of feedback have been
pioneered [43]. Particularly, the experiment of Barbero
and Grosse-Wentrup [43] showed that for an optimal
feedback design in BCIs, the current skill level of the
participants must be taken into account. In this study,
the investigators aimed to study how BCI-performance
is affected by biasing (i.e., no bias, strong positive bias,
weak positive bias, strong negative bias, and weak
negative bias) the belief participants have on their level
of control over the BCI system. The authors noted that
participants capable of modulating their sensorimotor
rhythms (SMR) to some level were impeded by
inaccurate feedback, whereas participants normally
performing on or close to chance level may essentially
benefit from an incorrect belief on their performance
level.
Neuper et al. [44] recently demonstrated that
continuous visual BCI feedback clearly modulates
sensorimotor EEG rhythms, however, no significant
differences between feedback conditions were found
when a realistic feedback (i.e., a grasping hand) and an
abstract feedback (i.e., a moving bar) were applied to
the subjects. The researchers observed that when the
feedback provided equivalent information on both the
continuous and final outcomes of mental actions, the
presentation form (i.e., abstract versus realistic) did not
influence the performance, that is, no difference
between sessions. It seems that there was no influence
of the feedback on task performance, probably by a
short training period or because the participants started
Progress and Prospects in EEG-Based Brain-Computer Interface
at a high performance level. Moreover, in both
conditions, the feedback stimuli seemed to become
directly associated with the action goal during online
control, and therefore, were able to enhance the
desired electrophysiological signals for subjects
accurately performing the task. Within this context,
such findings have a practical interest for classifier
development and BCI use for motor restoration.
EEG-BASED BCIs
In a typical EEG-based BCI setting, subjects are
exposed to stimuli or are required to execute mental
tasks while their cortical activity is being recorded by
EEG. Within this context, the relevant EEG features
extracted are then fed back to users by a so-called
closed-loop BCI. Specific features of the EEG are thus
regulated by BCI users. There are several other non-
invasive methods of monitoring brain activity, for
instance, transcranial magnetic stimulation [45, 46],
functional magnetic resonance imaging (fMRI) and
magnetoencephalography (MEG) [7, 19, 47-49];
however, we shall focus on the EEG technique. The
EEG-based BCI can be subdivided into groups
according to the electrophysiological signals used
(Figure 2). With this in mind, the following paragraphs
Figure 2: Main types of electrophysiological signals of EEG-based BCI.
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 31
will provide a short description of the neurophysiology
and of the use of these features for EEG-based BCI.
Slow Cortical Potential (SCP)
SCPs were among the first signals to be used to
drive a BCI system and can be interpreted as an
evoked response, time- and phase-locked to an event.
Because of this, averaging repeated signals will
increase the signal-to-noise ratio [50]. Users have to
produce positive or negative SCP shifts as compared
with the baseline; yet, these shifts must be above or
below the predefined threshold to be considered
negative or positive. A negative SCP shift represents
higher excitability; in contrast to the positive SCP shift,
which reflects reduced excitability or even inhibition [6].
With substantial training, control of SCPs to produce
positive or negative voltage shifts can be learned and
used for basic word processing and other simple
control tasks, such as accessing the internet. For
instance, in 1999, the very first verbal message ever
communicated with a brain-computer interface was
seen as a message on screen: written exclusively by
regulation of the EEG. In this study, two ‘locked-in’
patients with ALS were operant conditioned with direct
positive feedback selecting letters from computer-
32 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
presented letter strings using learned voluntary
decrease of SCPs. However, this feat required an
extensive training period consisting of up to 327
sessions [33].
Sensorimotor Rhythms
Sensorimotor rhythms (SMR) refer to localized
frequencies in the
-rhythm (11-13 Hz) and lower ß-
rhythm (15-18 Hz) bands EEG activity, which can be
recorded over primary sensorimotor cortex (SM1) [6,
29].
When certain tasks are performed, it is possible to
observe changes in on-going EEG activity, the so-
called event-related desynchronization (ERD) and
event-related synchronization (ERS). These describe
transient changes in on-going oscillatory EEG activity,
with ERD meaning a relative power decrease and ERS
meaning a relative power increase in specific spectral
components over defined cortical areas [10, 29]. SMR
decreases or desynchronizes with movement,
preparation for movement or MI, and increases or
synchronizes in the post-movement period [29, 51].
ERD and ERS are examples of induced responses,
occurring as a result of changes in the oscillatory
behavior of a group of neurons. These responses are
not phase-locked, but the power is time-locked to the
stimulus, meaning that the power in specific frequency
bands must be calculated before averaging across
trials [50]. This process is possible only due to a
different approach specifically designed to obtain
spatial filters called “method of common spatial
patterns” (CSP). The CSP technique [52] allows one to
determine spatial filters that maximize the variance of
signals of a certain condition (e.g., imagination of the
right-hand movement) and at the same time minimize
the variance of signals of another condition (e.g.,
imagination of the left-foot movement). Because the
variance of bandpass filtered signals is equal to the
bandpower, CSP filters are well suited to discriminate
mental states or movement-related patterns
characterized by ERD/ERS changes. Such changes
are based on a decomposition of the raw EEG signals
into spatial patterns, which are extracted from two
populations of single trial EEG, maximizing the
difference between them [53, 54].
For this process to be useful, the input to the
algorithm must be represented in such a way that
class-dependent changes in the signal are reflected in
a change in signal variance. In the case of ERD in MI,
this goal can be achieved by applying a zero-phase-
Machado et al.
distortion bandpass filter that captures the part of the
spectrum in which SMRs are expressed. The variance
of the filtered signal, which has zero mean, is a
measure of amplitude in the chosen band, i.e.,
-
rhythm (11-13 Hz) and lower ß-rhythm (15-18 Hz)
bands. As such, his approach has been successfully
applied in BCI systems [55, 56], in which CSP filters
are calculated individually for each subject on the data
of a calibration measurement.
Within this context, results from BCI studies have
shown that people can learn to control
-rhythm (11-13
Hz) or ß-rhythm (15-18 Hz) amplitudes in the absence
of any movement or sensation and can use this control
to move a cursor to select letters or icons on a screen
or to operate a simple orthotic device. For instance,
ALS patients were trained to learn to modulate SMR
receiving feedback, e.g. controlling cursor movement
on a computer screen in one or two dimensions. During
each trial of one-dimensional control, patients were
presented with a target consisting of a red vertical bar
that occupied the top or bottom half of the right side of
the screen and a cursor on the left side. The patient’s
task was to move the cursor so that it hit the target. As
a result, low SMR amplitude during MI moved the
cursor to the bottom bar, and high SMR amplitude (i.e.,
thinking of nothing in particular) moved the cursor
toward the top bar [57].
Visual Event-Related Potential (P300)
The evoked response that is used most often for
BCI is the P300 [5, 6, 18, 58]. The P300 response can
be interpreted as an electrical potential recorded from
the nervous system following the presentation of a
stimulus. For instance, P300 is typically seen when
subjects are required to attend to rare stimuli presented
within a stream of frequent standard stimuli, an
experimental design referred to as an oddball
paradigm. The P300 appears as a positive deflection
roughly 300 ms after stimulus presentation and is
related to the amount of attention to the stimulus. Thus,
in the context of BCI, if a set of distinct visual [59] or
auditory [18] stimuli are presented sequentially; the
P300 can indicate the stimulus selected by the subject
(serving as an ‘oddball’ by being the stimulus the
subject highlights through attention).
In this way, the P300 enables a BCI to convey the
subject’s request. The P300 has been used repeatedly
for speller applications in which letters, numbers, or
other visual stimuli are arranged in a matrix, and the
rows and columns of the matrix flash in rapid
Progress and Prospects in EEG-Based Brain-Computer Interface
succession while the user focuses attention on the item
to be selected. Only the row and column that contain
the specific item will generate a P300 potential. For
instance, Donchin and colleagues [60] presented the
user with a 6 x 6 matrix of characters. The rows and
columns in this matrix flashed successively and
randomly at a rate of eight flashes per second. The
user selected a character by focusing attention on it
and counting how many times it flashed. The row or
column that contained this character evoked a P300
response, whereas the others did not. After averaging
a number of responses, the computer could determine
the character’s row and column (as the row/column
with the highest P300 amplitude) and, thus, the desired
character. Nevertheless, to what extent the accuracy of
such a BCI requires eye movements (overt attention) or
whether it is also feasible for targets in the visual
periphery (covert attention) was still unclear, especially
with regard to how the BCI visual design could be
improved to meet the peculiarities of peripheral vision
(e.g., low spatial acuity and crowding). With this in
mind, Treder and Blankertz [61] investigated 13 healthy
subjects performing a copy-spelling task in which
subjects had to count the number of the intensifications
of the target symbol while their
EEG activity and eye movements were recorded
simultaneously. Participants had to fixate a dot in the
center of the screen and allocate their attention to a
target in the visual periphery. In addition, the effect of
visual speller design was examined, by comparing the
symbol Matrix to an ERP-based Hex-o-Spell (i.e., two-
levels speller consisting of six discs arranged on an
invisible hexagon). Thus, the investigators noted an
advantage (i.e., less errors, larger ERP amplitude
modulation, and better classification) of overt attention
over covert attention, and an advantage of the Hex-o-
Spell over the Matrix. Using overt attention, early ERP
components were enhanced by attention in comparison
with covert attention, while later ERP components were
enhanced for both spellers and were modulated when
using the Hex-o-Spell but not when using the Matrix.
As a result, classifiers rely mainly on early evoked
potentials in overt attention, in contrast to covert
attention in which classifiers rely mainly on later
cognitive components. Within this context, both overt
and covert attention can be used to drive an ERP-
based BCI; however, performance is clearly lower for
covert attention. Moreover, the Hex-o-Spell out-
performed the Matrix, especially when eye movements
were not permitted; indicating that performance can be
increased if one accounts for peculiarities of peripheral
vision.
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 33
Steady-State Visual Evoked Potential (SSVEPs)
Another type of evoked response is the SSVEP,
which refers to a periodic signal that is modulated with
a known periodic pattern. This pattern can be traced in
measurements of brain activity recorded at occipital
brain areas, when the subject is focusing on a flickering
visual stimulus at different frequencies [62], each
encoding a different command (up/down/right/left);
those signals are then processed to move the cursor.
The oscillation frequency of the SSVEP is driven by
that of the attended stimulus, which allows the subject
to select one out of several simultaneously presented
stimuli flickering with different frequencies. The power
and phase of the signal can be influenced by selective
attention by the subject, thus providing a suitable task
for a BCI [63].
Kelly et al. [64] investigated classification
accuracies when users were not required to focus gaze
on the flickering targets but rather on a fixation cross
between two targets (covert attention). A decrease in
accuracy was observed from about 95% when the
targets were fixated directly, to about 70% in the covert
attention condition. Therefore, it seems that this
paradigm might be used by ‘locked-in’ patients who
cannot direct gaze, albeit with reduced accuracy.
Moreover, a novel independent SSVEP-BCI system,
based on covert non-spatial visual selective attention,
was created by Zhang et al. [65]. In this experiment,
healthy subjects engaged in a task of superimposed
illusory surfaces perception induced by two sets of dots
with different colors rotating in opposite directions. An
online BCI system using attentional modulation of
SSVEPs was employed and a 3-day online training
was carried out. As a result, in 8 of 18 subjects, a
general improvement of control accuracy with training
was observed. The findings showed that SSVEP-based
BCI paradigms are useful for paralyzed patients with
substantial head or ocular motor impairments by
employing covert attention shifts instead of changing
gaze direction.
To summarize, available P300-, SSVEP-, SMR-, or
SCP-based BCIs rely mainly on visual stimuli and
visual feedback. Although, these BCIs do not depend
on voluntary muscle control (i.e., eye movements), they
nevertheless do need the users to be able to maintain
gaze [66-68]. In addition, the BCIs certainly depend to
some degree on normal brain functions. All things
considered, the electrophysiological signals reflect a
combination of cortical and subcortical functions. For
instance, impairments of cortex (e.g., ALS and stroke),
34 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
basal ganglia, or other subcortical areas that interact
with cortex (e.g., cerebral palsy) or loss of afferent
input (e.g., brainstem stroke or spinal cord injury) could
affect the user’s ability to achieve control of SSVEP,
P300, SMR, or SCP. Thus, the ability to use the BCI’s
systems and the best alternative among the different
types of them, are likely to differ among users [67].
Moreover, people who are severely disabled might not
have the visual acuity or gaze stability needed to see
the visual stimuli associated with BCI use, particularly if
the stimuli change rapidly. Thus, BCI systems that use
auditory rather than visual stimuli might be preferable,
or even crucial, for some users [69, 70].
TYPES OF BCI
BCIs have been categorized in many ways in recent
years. Although not all BCI researchers use the same
terminology, some subdivisions of BCIs are used. With
this in mind, we will present the most common
subdivisions (i.e., dependent, independent,
synchronous and asynchronous BCIs) of BCI and their
functions. These subdivisions have so far been studied
mostly in mutually exclusive fashion, but future BCIs
may combine some or all of the above in one system,
the called hybrid BCI.
Dependent vs Independent BCIs
BCI systems can be classified as dependent or
independent [5, 71]. Dependent BCI requires some
muscle control to produce the neural activity used for
communication or control [5, 71], such as the SSVEP
which usually depends on gaze control and thus on
muscle activity. This type of dependent BCI presents
the user who is paralyzed with a matrix of letters that
flash one at a time, and the user selects a specific
letter by looking directly at it so that the visual evoked
potential (VEP) is recorded over visual cortex when
that letter flashes is much larger that the VEPs
produced when other letters flash.
Therefore, the relevant signal is coming from the
EEG but is due to the mechanisms of sight.
Specifically, the brain’s output channel is the EEG, but
the generation of its signal depends on gaze direction
and therefore on extraocular muscles [16]. This
technique is essentially an alternative method for
detecting messages carried in the brain’s normal output
pathways: in the present example, gaze direction is
detected by monitoring EEG rather than by monitoring
eye position directly. Although dependent BCIs may be
of use for a variety of applications, individuals without
motor control may not be able to use them [72].
Machado et al.
Independent or ‘pure’ BCIs do not require any
muscle control. In other words, this type is entirely free
from the physiological output pathways of the brain, as
the relevant signal is not generated by propagating
signals along peripheral nerves, muscles, or other
physiological outputs. The user may simply imagine or
feel a limb moving without the real movement. In this
way, the brain signals would then reveal the user’s
mental task regardless of control of the limb [6]. For
example, typical P300-based BCIs presents the user
with a matrix of letters that flash one at a time, and the
user selects a specific letter by producing a P300-
evoked potential when that letter flashes. This
response appears to be independent of muscle activity,
and the generation of the EEG signal depends mainly
on the user’s intent, not on the precise orientation of
the eyes. Furthermore, for people with the most severe
neuromuscular disabilities, who may lack all normal
output channels, including extraocular muscle control,
independent BCIs, are likely to be more useful [73]. On
the other hand, less severely disabled patients would
benefit from dependent BMIs as well. Thus, which type
one chooses depends on the application scenario and
the population one targets.
Synchronous vs Asynchronous BCIs
The mode of operation determines when the user
performs a mental task and thereby intends to transmit
a message. In principle, there are two different modes
of operation [38]. The first is computer-driven, meaning
that operation is externally paced by a cue (cue-based
or synchronous BCI). In the case of a synchronous
BCI, the mental task has to be performed in predefined
time windows following a visual or auditory cue
stimulus [74]. The time periods during which the user
can exert control, for example, by ‘producing’ a specific
mental MI, are determined by the system, and the
processing of the data is limited to these fixed periods.
The majority of work in current BCI research is based
on this synchronous mode [38]. This mode has been
applied to control locomotion in a virtual reality
environment (i.e. a virtual street). In this case, forward
walking was controlled by imagination of leg/foot
movements and to stop the walking, right hand
movement was imagined [74].
The second mode is user-driven, internally paced
without using any cue (uncued or asynchronous BCI).
The asynchronous BCI requires a continuous analysis
and data-driven feature extraction from the recorded
signals. In this way, the system must be able to
recognize when an intended mental state occurs.
Progress and Prospects in EEG-Based Brain-Computer Interface
Asynchronous operation is thus more complex than a
synchronous BCI. On the other hand, the former allows
a more natural form of communication because the
user can decide when a command should be sent [38,
74]. Asynchronous BCI has been used to control, via
functional electrical stimulation, hand grasp function in
tetraplegic patients, using the dynamics of brain
oscillations caused by foot MI [29, 75]. Other
asynchronous BCI studies based on hand or finger MI
have been used for cursor control in tetraplegic
patients [76] and spelling device control in healthy
patients [77, 78].
Although this kind of BCI provides more freedom,
the system may be more vulnerable to the
phenomenon of unrelated activity being interpreted as
a message or command. If a command could come at
any moment, the brain signals are analyzed for
possible commands continuously. Continuous feature
extraction thus increases the possibility of ‘false
positives’; erroneous mental state detection, if only
because more brain signals are analyzed. These
issues have fostered the exploration of methods for
turning a BCI on or off with brain activity. Although
more problematic, asynchronous operation can be
developed further because for real-world applications,
when the user needs full control over timing and speed
of BCI operation, the asynchronous communication
mode is highly desirable [38, 79].
Hybrid BCIs
Recently, however, researchers introduced a novel
combination of tasks that could inspire BCI systems
and that seem to be more accurate than conventional
BCIs, especially for users who cannot attain accuracy
adequate for effective communication [80, 81]. This
novel system is called hybrid brain-computer interface
(BCI), composed of two BCIs or at least one BCI and
another system. A hybrid BCI, like any BCI, must fulfill
the following criteria—the device must rely on signals
recorded directly from the brain, there must be at least
one recordable brain signal that the user can purposely
modulate to effect a goal-directed behavior, real time
processing, and the user must get feedback. In view of
that, Pfurtscheller et al. [81] described BCIs that
classify ERD/ERS, and SSVEP that can either process
their inputs concurrently or operate two systems
successively, wherein the first system can act as a
“brain switch”, for instance, the ERD. Within this
context, Allison et al. [80] evaluated the feasibility of
combining two mental tasks that simulated a
simultaneous hybrid BCI. Each trial of the experiment
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 35
began with an arrow pointing to the left or right sides,
indicating that the subject should perform a left or right
MI task. In the ERD condition, subjects imagined left
and right hand movement. In the SSVEP condition, the
left arrow cued the subject to focus attention on the left
LED that flickered at 8, while the right arrow cued the
subject to focus attention on the right LED that flickered
at 13 Hz. In the hybrid condition, the left arrow cued the
subject to simultaneously imagine left hand movement
and focus attention on the left LED, while the right
arrow cued the subject to simultaneously imagine right
hand movement and focus attention on the right LED.
The classification accuracy was highest in the hybrid
condition, although without statistical significance. In
addition, in both ERD and SSVEP conditions, some
subjects could not attain proficiency, resulting in low
classification accuracy for an effective communication.
Yet, when these subjects were not proficient with either
the ERD or SSVEP BCIs, they were usually proficient
with the other type of BCI. Taking these findings into
account, it seems a subject who could not use an ERD
BCI might attain proficiency with an SSVEP BCI and
vice versa. With regard to the hybrid condition, the
number of subjects who cannot attain proficiency was
significantly lower than in the ERD condition, whereas
no significant difference was observed in the number of
subjects that could not attain proficiency in the SSVEP-
hybrid comparison or the ERD-SSVEP comparison.
These findings validate the hybrid BCI and suggest that
subjects who cannot use a BCI (i.e., ERD or SSVEP
BCI) ought to consider switching to a new BCI system,
particularly a hybrid BCI. Therefore, a hybrid BCI using
two tasks might be more accurate than a conventional
BCI using either task (i.e., MI or SSVEP BCI). For this
reason, the subjects’ proficiency with both BCIs might
be combined to increase information output by
improving the accuracy, reducing the selection time,
and/or increasing the number of possible commands.
CLINICAL APPLICATION OF EEG-BASEDBCI
BCI technologies have wide range of potential
applications, ranging from simple to complex. Simple
BCI applications have been validated in the laboratory
but are in limited clinical use. These methods include
systems for answering “yes” or “no” to questions,
managing basic control of the user’s environment (e.g.,
lights and temperature), controlling a television, or
opening and closing a hand orthosis. These simple
systems can be configured for basic word processing,
sending e-mails, accessing the internet, or operating a
motorized wheelchair. In relation to these issues, some
people who are severely disabled currently use EEG-
36 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
based BCI systems for important purposes in their daily
lives [82] These systems might also support more
complex applications, such as the operation of a
robotic arm or a neuroprosthetic limb that provides
multi-dimensional movement to an otherwise paralyzed
limb, e.g., for hemiparetic stroke patients. Although
many efforts are focusing on developing invasive BCI
systems for these complex uses [83], non-invasive
EEG-based BCIs might also serve these purposes [14,
35, 84]. Naturally, non-invasive applications would,
once they are fully functional, be preferable. Such BCI
applications might enable mainly people who are
almost totally paralyzed (i.e., “locked-in” or
“tetraplegia”) to have a higher quality of life that can
also be productive.
In addition to assisted communication, BCIs
involving operant learning of EEG slow cortical
potentials and sensorimotor rhythm were demonstrated
to be successful in drug-resistant focal epilepsy and
attention deficit disorder [6, 71]. The future importance
of such BCI applications will depend on their
capacities, practicality, and reliability, their acceptance
by particular groups of users, and on the extent to
which they have substantial advantages over
conventional assistive technology. So, given this
promise of BCI in patient populations, we should
evaluate what has actually been achieved in patient
groups thus far.
Amyotrophic Lateral Sclerosis (ALS)
Communication with BCI
In 2006, Pfurtscheller et al. [29] reported a case of a
patient (60 years of age, male) who had suffered from
amyotrophic lateral sclerosis (ALS) since 1998 and had
lost almost all voluntary muscle control. After training in
a basket paradigm, in which the goal was to hit a target
positioned at the bottom of a screen by controlling the
horizontal position of a ball falling at a constant speed
from the top of the screen, he learned to reliably
produce two different EEG patterns (imagination of left
and right hand movements). The first one was
characterized by a broad-banded ERD and the other
one by a narrow-banded ERS in the form of induced
10-Hz oscillations (ERS), similar to able-bodied
subjects.
The BCI-control achieved enabled the patient to use
a 2-class virtual keyboard based on spelling. Birbaumer
et al. [33], and Kübler et al. [68, 85] reported on several
ALS patients with several degrees of impairment (no
eye and sphincter movements or other form of motor
Machado et al.
activity present or paralysis of legs or arms) trained
with SCP, P300 and the SMR BCI systems. Although
seven patients were excluded after the training
because they were not able to use the BCI system,
most patients achieved significant control over their
brain activity and were able to select letters and write
words with one of the three BCI systems. Six patients
achieved complete independence of communication
with the BCI. From the studies summarized, which of
the three non-invasive BCIs is the most promising is
not clear. With this in mind, eight severely paralyzed
patients with ALS were trained with SCP, P300, and
SMR BCIs in a balanced crossover-within-subject
design. In other words, the patients were confronted
with various types of BCI to provide them with a system
that works best for the individual patient and to
determine whether one BCI type would be superior to
another. Each training block of each BCI type lasted 20
sessions. SCP and SMR BCI improved control of brain
activity within this short training period but were not
good enough to select letters (70% minimum success
rate). The fastest acquisition and fastest spelling rates
were achieved with the P300 BCI, with more than 70%
accuracy in using the spelling system [86].
Paralysis from Spinal Cord Lesions
In 2000, Pfurtscheller et al. [40] reported a young
tetraplegic patient (i.e., 22 years old) who had suffered
from a traumatic spinal cord injury since 1998 but after
and extensive training to increase and decrease central mu-
rhythms was able to control an electrostimulation
device (FES) applied to the hand muscles. The patient
was thus able to grasp a glass and bring it to his mouth
after he had learned with feedback and reward over a
period of 4 months to regulate his mu-rhythm. Three
years ago, Pfurtscheller et al. [75] studied with spinal
cord injury (SCI) patients a self-paced MI-controlled
operation of a neuroprosthesis. The grasp function of
the left hand of the first patient (29 year, male, SCI at
level C5) was restored with FES using surface
electrodes. During a 4-month ERD-BCI-training period,
the patient learned to induce 17-Hz oscillations by
means of foot MI that became sufficiently dominant to
control the FES device. The second patient (42 years
old, male, SCI sub-C5) had a freehand system
implanted in his right hand and arm. Within 3 days of
feedback training, he learned to reliably induce an ERD
pattern during left-hand MI and thus to generate a
binary control signal. In this case, the self-paced BCI
system emulated the shoulder joystick that is usually
used to operate the freehand system. With the BCI-
controlled freehand system, the patient successfully
Progress and Prospects in EEG-Based Brain-Computer Interface
executed parts of a hand-grasp performance test.
These experiments may serve as evidence for the
feasibility of controlling an implantable neuroprosthesis
via a noninvasive BCI. Those findings provide the basis
for the development of ‘thought’-controlled
neuroprostheses that might help patients with severe
paralysis to regain control over their bodies.
Stroke Rehabilitation
In 2009, Ang et al. [87] investigated the effects of a
SMR-BCI system based on MI for upper limb robotic
rehabilitation in hemiparetic stroke patients. Patients
were randomly divided into two groups; MI-BCI robotic
rehabilitation (n = 8) and standard robotic rehabilitation
(n = 10). Each subject underwent 12 sessions of 1-hour
rehabilitation for 4 weeks. Significant gains in Fulg-
Meyer scale scores were observed in both groups at
postrehabilitation and 2-month post-rehabilitation. The
experimental group yielded a higher 2-month
postrehabilitation gain than the control (6.0 versus 4.0)
but no significance was found. Nevertheless, among
subjects with positive gain (n = 6, n= 7, respectively),
the initial difference of 2.8 between the groups
increased to a significant 6.5 after adjustment for age
and gender. Therefore, this study provides evidence
that BCI-driven robotic rehabilitation is effective in
restoring motor control for hemiparetic stroke patients.
A similar procedure was used by Broetz et al. [88] to
reduce focal impairments for a hemiplegic stroke
patient with no active finger extension. Training with an
SMR-BCI combined with a daily life–oriented physical
therapy protocol was applied. The SMR-BCI was used
to drive an orthosis and a robot affixed to the patient’s
affected upper limb through MI of grasp movements,
which enabled him to move the paralyzed arm and
hand driven by voluntary modulation of
-rhythm
activity. The goal-directed physical therapy training
(i.e., grasp movements and walking) was performed
over a period of 1 year, during which patients
completed 3 training blocks. In this way, the upper limb
motor functions and brain reorganization were
repeatedly assessed along the study using
sensorimotor scales and tests (e.g., Fugl-Meyer, Wolf
Motor Function, and Modified Ashworth) and using
fMRI and MEG. The ability of upper limb movements
improved significantly by 46.6%, and its improvement
was associated with increased
-rhythm activity in the
ipsilesional motor cortex. The findings indicate that the
BCI training combined with goal-directed physical
therapy protocol may improve the motor functions of
chronic stroke patients despite visible initial paralysis.
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 37
Downregulation of Cortical Excitation in Epilepsy
and Attention Regulation in ADHD
In 1977, Sterman [6] reported that humans were
trained to increase SMRs and after extensive training,
some patients with drug-resistant epilepsy were able to
demonstrate seizure reduction and remission. In a few
studies [89, 90], patients with focal epileptic seizures
were trained to down-regulate cortical excitation by
rewarding them for positive cortical potentials and
perception of SCP changes. After 35 sessions of
training, some of these patients gained close to 100%
control over their SCPs and consequently there was a
strong decrease of seizure activity. Clinically the BCI
training with focal intractable epilepsies was highly
successful in reducing seizure frequencies; on average
to half of the baseline and controls. One to two years of
follow-ups demonstrated stable improvements and
some patients remained seizure free. Only patients
with very high negative SCP amplitudes before training
did not profit from this SCP-BCI [89].
A similar procedure was applied in children with
ADHD [91, 92]. The studies showed that both learning
to induce an increase of central-frontal negativity of
SCPs and increase of SMRs and beta frequencies
improves the symptoms. These EEG changes were
associated with an improvement of ADHD symptoms
comparable with medication effects (i.e.,
Methylphenidate). No difference inefficacy was found
between SCP, beta frequency, or SMR training,
suggesting that the behavioral effects rely on the
convergence of control of different EEG activities on a
final common therapeutic pathway, possibly an
improvement in the general capacity to regulate
attention via brain regulation.
CONCLUSION AND FUTURE CONSIDERATIONS
EEG-based BCI proposes the development of
interfaces based on the interaction of neural networks
with artificial tools to restore motor control and full
mobility of the injured area. The cortical physiology that
corroborates the manner in which a human brain
encodes intentions is beginning to be understood and
will have a substantial affect in augmenting function in
those with various forms of motor disabilities, such as
post-stroke patients and tetraplegic people. As
research stretches beyond motor physiology, the field
of EEG-based BCI stands to further expand its
applications and the diversity of the clinical population
served. Given the rapid progression of EEG-based BCI
over the past years and the concomitant swift ascent of
38 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
computer processing speeds and signal analysis
techniques, this approach will yield new feasible clinical
applications in the near future.
ACKNOWLEDGEMENTS
All authors participated conducting the literature
review and drafted most of the manuscript. All authors
were equally involved in reading and approving the
final manuscript.
REFERENCES
[1] Isa T, Fetz EE, Müller KR. Recent advances in brain-
machine interfaces. Neural Netw 2009; 22: 1201-2.
http://dx.doi.org/10.1016/j.neunet.2009.10.003
[2] Mason SG, Bashashati A, Fatourechi M, Navarro KF, Birch
GE. A comprehensive survey of brain interface technology
designs. Ann Biomed Eng 2007; 35: 137-69.
http://dx.doi.org/10.1007/s10439-006-9170-0
[3] Millan J del R, Rupp R, Mueller-Putz G, Murray-Smith R,
Giugliemma C, Tanger-mann M, et al. Combining brain-
computer interfaces and assistive technologies: State-of-the-
art and challenges. Front Neuro-Prosthetics 2010; 4: 161.
[4] Wang W, Collinger JL, Perez MA, Tyler-Kabara EC, Cohen
LG, Birbaumer N, et al. Neural interface technology for
rehabilitation: exploiting and promoting neuroplasticity. Phys
Med Rehabil Clin N Am 2010; 21: 157-78.
http://dx.doi.org/10.1016/j.pmr.2009.07.003
[5] Wolpaw JR, Birbaumer N, McFarland DJ, Pfurtscheller G,
Vaughan TM. Brain-computer interfaces for communication
and control. Clin Neurophysiol 2002; 113: 767-91.
http://dx.doi.org/10.1016/S1388-2457(02)00057-3
[6] Birbaumer N, Cohen LG. Brain-computer interfaces:
communication and restoration of movement in paralysis. J
Physiol 2007; 579: 621-36.
http://dx.doi.org/10.1113/jphysiol.2006.125633
[7] Birbaumer N, Ramos Murguialday A, Weber C, Montoya P.
Neurofeedback and brain computer interface clinical
applications. Int Rev Neurobiol 2009; 86: 107-17.
http://dx.doi.org/10.1016/S0074-7742(09)86008-X
[8] Blankertz B, Losch F, Krauledat M, Dornhege G, Curio G,
Müller KR. The Berlin Brain-Computer Interface: accurate
performance from first-session in BCI-naive subjects. IEEE
Trans Biomed Eng 2008; 55: 2452-62.
http://dx.doi.org/10.1109/TBME.2008.923152
[9] Scherer R, Müller-Putz GR, Pfurtscheller G. Flexibility and
practicality graz brain-computer interface approach. Int Rev
Neurobiol 2009; 86: 119-23.
http://dx.doi.org/10.1016/S0074-7742(09)86009-1
[10] Dornhege G, Millan JR, Hinterberger T, McFarland D, Müller
K. Toward brain-computer interfacing. Cambridge, MA: The
MIT Press 2007.
[11] Pfurtscheller G, Brunner C, Schlögl A, Lopes da Silva FH. Mu
rhythm (de)synchronization and EEG single-trial classification
of different motor imagery tasks. NeuroImage 2006; 31: 153-
9.
http://dx.doi.org/10.1016/j.neuroimage.2005.12.003
[12] Wolpaw JR. Brain-computer interfaces (ICCs) for
communication and control: a mini-review. Suppl Clin
Neurophysiol 2004; 57: 607-13.
http://dx.doi.org/10.1016/S1567-424X(09)70400-3
[13] Krusienski DJ, Wolpaw JR. Brain-computer interface
research at the wadsworth center developments in
Machado et al.
noninvasive communication and control. Int Rev Neurobiol
2009; 86: 147-57.
http://dx.doi.org/10.1016/S0074-7742(09)86011-X
[14] Wolpaw JR, McFarland DJ. Control of two-dimensional
movement signal by a noninvasive brain-computer interface
in humans. Proc Natl Acad Sci USA 2004; 101: 17849-54.
http://dx.doi.org/10.1073/pnas.0403504101
[15] Birbaumer N. Slow cortical potentials: Plasticity, operant
control, and behavioral effects. Neuroscientist 1999; 5: 74-8.
http://dx.doi.org/10.1177/107385849900500211
[16] Birbaumer N. Breaking the silence: brain-computer interfaces
(BCI) for communication and motor control.
Psychophysiology 2006; 43: 517-32.
http://dx.doi.org/10.1111/j.1469-8986.2006.00456.x
[17] Birbaumer N, Kubler A, Ghanayim N, Hinterberger T,
Perelmouter J, Kaiser J, et al. The thought translation device
(TTD) for completely paralyzed patients. IEEE Trans Rehabil
Eng 2000; 8: 190-3.
http://dx.doi.org/10.1109/86.847812
[18] Sellers EW, Donchin E. A P300-based brain-computer
interface: Initial tests by ALS patients. Clin Neurophysiol
2006; 117: 538-48.
http://dx.doi.org/10.1016/j.clinph.2005.06.027
[19] Weigelt S, Muckli L, Kohler A. Functional magnetic
resonance adaptation in visual neuroscience. Rev Neurosci
2008; 19: 363-80.
http://dx.doi.org/10.1515/REVNEURO.2008.19.4-5.363
[20] Buttfield A, Ferrez PW, Millan J del R. Towards a robust BCI:
Error recognition and online learning. IEEE Trans Neural Sys
Rehabil Eng 2006; 14: 164-8.
http://dx.doi.org/10.1109/TNSRE.2006.875555
[21] Lu S, Guan C, Zhang H. Unsupervised brain computer
interface based on inter-subject information and online
adaptation. IEEE Trans Neural Systems Rehabil Eng 2009;
17: 135-45.
http://dx.doi.org/10.1109/TNSRE.2009.2015197
[22] Millan JR. Adaptive brain interfaces. Comm ACM 2003; 46:
74-80.
http://dx.doi.org/10.1145/636772.636773
[23] Millan J del R, Renkens F, Mourino J, Gerstner W.
Noninvasive brain-actuated control of a mobile robot by
human EEG. IEEE Trans Biomed Eng 2004; 51: 1026-33.
http://dx.doi.org/10.1109/TBME.2004.827086
[24] Shenoy P, Krauledat M, Blankertz B, Rao RPN, Müller KR.
Towards adaptive classification for BCI. J Neural Eng 2006;
3: 13-23.
http://dx.doi.org/10.1088/1741-2560/3/1/R02
[25] Vidaurre C, Blankertz B. Towards a cure for BCI illiteracy.
Brain Topogr 2010; 23: 194-8.
http://dx.doi.org/10.1007/s10548-009-0121-6
[26] Pfurtscheller G, Neuper C, Guger C, HarkamW, Ramoser H,
Schlögl A, et al. Current trends in Graz brain-computer
interface (BCI) research. IEEE Trans Rehabil Eng 2000; 8:
216-9.
http://dx.doi.org/10.1109/86.847821
[27] Curran E, Sykacek P, Stokes M, Roberts SJ, Penny W,
Johnsrude I, et al. Cognitive tasks for driving a brain-
computer interfacing system: a pilot study. IEEE Trans
Neural Syst Rehabil Eng 2004; 12: 48-54.
http://dx.doi.org/10.1109/TNSRE.2003.821372
[28] Naito E, Sadato N. Internal simulation of expected sensory
experiences before movements get started. Rev Neurosci
2003; 14: 387-99.
http://dx.doi.org/10.1515/REVNEURO.2003.14.4.387
[29] Pfurtscheller G, Müller-Putz GR, Schlögl A, Graimann B,
Scherer R, Leeb R, et al. 15 Years of BCI Research at Graz
University of Technology: Current Projects. IEEE Trans
Neural Syst Rehabil Eng 2006; 14: 205-10.
http://dx.doi.org/10.1109/TNSRE.2006.875528
Progress and Prospects in EEG-Based Brain-Computer Interface
[30] Blankertz B, Dornhege G, Krauledat M, Müller KR, Curio G.
The noninvasive Berlin brain-computer interface: fast
acquisition of effective performance in untrained subjects.
Neuroimage 2007; 37: 539-50.
http://dx.doi.org/10.1016/j.neuroimage.2007.01.051
[31] Pfurtscheller G, Neuper C. Future prospects of ERD/ERS in
the context of brain-computer interface (BCI) developments.
Prog Brain Res 2006; 159: 433-7.
http://dx.doi.org/10.1016/S0079-6123(06)59028-4
[32] McFarland DJ, Krusienski DJ, Sarnacki WA, Wolpaw JR.
Emulation of computer mouse control with a noninvasive
brain–computer interface. J Neural Eng 2008; 5: 101-10.
http://dx.doi.org/10.1088/1741-2560/5/2/001
[33] Birbaumer N, Ghanayim N, Hinterberger T, Iversen I,
Kotchoubey B, Kübler A, et al. A spelling device for the
paralysed. Nature 1999; 398: 297-8.
http://dx.doi.org/10.1038/18581
[34] Chatterjee A, Aggarwal V, Ramos A, Acharya S, Thakor NV.
A brain-computer interface with vibrotactile biofeedback for
haptic information. J Neuroeng Rehabil 2007; 4: 40.
http://dx.doi.org/10.1186/1743-0003-4-40
[35] Middendorf M, McMillan G, Calhoun G, Jones KS. Brain-
computer interfaces based on the steady-state visual-evoked
response. IEEE Trans Rehabil Eng 2000; 8: 211-3.
http://dx.doi.org/10.1109/86.847819
[36] Müller-Putz GR, Scherer R, Pfurtscheller G, Rupp R. EEG-
based neuroprosthesis control: a step towards clinical
practice. Neurosci Lett 2005; 382: 169-74.
http://dx.doi.org/10.1016/j.neulet.2005.03.021
[37] Müller-Putz GR, Scherer R, Pfurtscheller G, Rupp R. Brain-
computer interfaces for control of neuroprostheses: From
synchronous to asynchronous mode of operation. Biomed
Tech (Berl) 2006; 51: 57-63.
http://dx.doi.org/10.1515/BMT.2006.011
[38] Neuper C, Müller-Putz GR, Scherer R, Pfurtscheller G. Motor
imagery and EEG based control of spelling devices and
neuroprostheses. Prog Brain Res 2006; 159: 393-409.
http://dx.doi.org/10.1016/S0079-6123(06)59025-9
[39] Lauer, R, Peckham, P, Kilgore, K, Heetderks, W.
Applications of cortical signals to neuro-prosthetic control: A
critical review. IEEE Trans Rehabil 2000; 8: 205-8.
http://dx.doi.org/10.1109/86.847817
[40] Pfurtscheller G, Guger C, Mueller G, Krausz G, Neuper C.
Brain oscillations control hand orthosis in a tetraplegic.
Neurosci Lett 2000; 292: 211-4.
http://dx.doi.org/10.1016/S0304-3940(00)01471-3
[41] Sinkjaer T, Haugland M, Inmann A, Hansen M, Nielsen KD.
Biopotentials as command and feedback signals in functional
electrical stimulation systems. Med Eng Phys 2003; 25: 29-
40.
http://dx.doi.org/10.1016/S1350-4533(02)00178-9
[42] Cincotti F, Kauhanen L, Aloise F, Palomäki T, Caporusso N,
Jylänki P, et al. Vibrotactile feedback for brain computer
interface operation. Comput Intell Neurosci 2007; 48937.
[43] Barbero A, Grosse-Wentrup M. Biased feed back in brain-
computer interfaces. J Neuroeng Rehabil 2010; 7: 34.
http://dx.doi.org/10.1186/1743-0003-7-34
[44] Neuper C, Scherer R, Wriessnegger S, Pfurtscheller G.
Motor imagery and action observation: modulation of
sensorimotor brain rhythms during mental control of a brain-
computer interface. Clin Neurophysiol 2009; 120: 239-47.
http://dx.doi.org/10.1016/j.clinph.2008.11.015
[45] Bonnard M, de Graaf J, Pailhous J. Interactions between
cognitive and sensorimotor functions in the motor
cortex:evidence from the preparatory motor sets anticipating
a perturbation. Rev Neurosci 2004; 15: 371-82.
http://dx.doi.org/10.1515/REVNEURO.2004.15.5.371
Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 39
[46] Tyc F, Boyadjian A. Cortical plasticity and motor activity
studied with transcranial magnetic stimulation. Rev Neurosci
2006; 17: 469-95.
http://dx.doi.org/10.1515/REVNEURO.2006.17.5.469
[47] Birbaumer N, Weber C, Neuper C, Buch E, Haapen K,
Cohen L. Physiological regulation of thinking: brain-computer
interface (BCI) research. Prog Brain Res 2006; 159: 369-91.
http://dx.doi.org/10.1016/S0079-6123(06)59024-7
[48] Sitaram R, Caria A, Birbaumer N. Hemo-dynamic brain-
computer interfaces for communication and rehabilitation.
Neural Netw 2009; 22: 1320-8.
http://dx.doi.org/10.1016/j.neunet.2009.05.009
[49] Sorger B, Dahmen B, Reithler J, Gosseries O, Maudoux A,
Laureys S, et al. Another kind of ‘BOLD Response’:
answering multiple-choice questions via online decoded
single-trial brain signals. Prog Brain Res 2009; 177: 275-92.
http://dx.doi.org/10.1016/S0079-6123(09)17719-1
[50] Pfurtscheller G, Lopes da Silva FH. Event-related EEG/MEG
synchronization and desynchronization: basic principles. Clin
Neurophysiol 1999; 110: 1842-57.
http://dx.doi.org/10.1016/S1388-2457(99)00141-8
[51] Toro C, Deuschl G, Thatcher R, Sato S, Kufta C, Hallett M.
Event-related desynchronization and movement-related
cortical potentials on the ECoG and EEG.
Electroencephalogr Clin Neurophysiol 1994; 93: 380.
http://dx.doi.org/10.1016/0168-5597(94)90126-0
[52] Fukunaga K. Inroduction to statistical pattern recognition.
2nd edition. San Diego, CA: Academic Press 1990.
[53] Koles ZJ, Soong ACK. EEG source localization:
implementing the spatiotemporal decomposition approach.
Electroencephalogr Clin Neurophysiol 1998; 107: 343-52.
http://dx.doi.org/10.1016/S0013-4694(98)00084-4
[54] Müller-Gerking J, Pfurtscheller G, Flyvbjerg H. Designing
optimal spatial filters for single-trial EEG classification in a
movement task. Clin Neurophysiol 1999; 110: 787-98.
http://dx.doi.org/10.1016/S1388-2457(98)00038-8
[55] Blankertz B, Dornhege G, Krauledat M, Müller KR,
Kunzmann V, Losch F, et al. The Berlin Brain-Computer
Interface: EEG-based communication without subject
training. IEEE Trans Neural Syst Rehabil Eng 2006; 14: 147-
52.
http://dx.doi.org/10.1109/TNSRE.2006.875557
[56] Guger C, Ramoser H, Pfurtscheller G. Real-time EEG
analysis with subject specific spatial patterns for a brain
computer interface (BCI). IEEE Trans Rehabil Eng 2000; 8:
447-56.
http://dx.doi.org/10.1109/86.895947
[57] Kübler A, Nijboer F, Mellinger J, Vaughan TM, Pawelzik H,
Schalk G, et al. Patients with ALS can use sensorimotor
rhythms to operate a brain-computer interface. Neurology
2005; 64: 1775-7.
http://dx.doi.org/10.1212/01.WNL.0000158616.43002.6D
[58] Piccione F, Giorgi F, Tonin P, Priftis K, Giove S, Silvoni S, et
al. P300-based brain-computer interface: Reliability and
performance in healthy and paralysed participants. Clin
Neurophysiol 2006; 117: 531-7.
http://dx.doi.org/10.1016/j.clinph.2005.07.024
[59] Hoffmann U, Vesin JM, Ebrahimi T, Diserens K. An efficient
P300-based brain-computer interface for disabled subjects. J
Neurosci Methods 2008; 167: 115-25.
http://dx.doi.org/10.1016/j.jneumeth.2007.03.005
[60] Donchin E, Spencer KM, Wijesinghe R. The mental
prosthesis: Assessing the speed of a P300-based brain-
computer interface. IEEE Trans Rehab Eng 2000; 8: 174-9.
http://dx.doi.org/10.1109/86.847808
[61] Treder MS, Blankertz B. (C)overt attention and visual speller
design in an ERP-based brain-computer interface. Behav
Brain Funct 2010; 6: 28.
http://dx.doi.org/10.1186/1744-9081-6-28
40 Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1
[62] Bakardjian H, Tanaka T, Cichocki A. Optimization of SSVEP
brain responses with application to eight-command Brain-
Computer Interface. Neurosci Lett 2010; 469: 34-8.
http://dx.doi.org/10.1016/j.neulet.2009.11.039
[63] Krusienski DJ, Allison BZ. Harmonic coupling of steady-state
visual evoked potentials. Conf Proc IEEE Eng Med Biol Soc
2008; 2008: 5037-40.
[64] Kelly SP, Lalor EC, Reilly RB, Foxe JJ. Visual spatial
attention tracking using high-density SSVEP data for
independent brain-computer communication. IEEE Trans
Neural Syst Rehab Eng 2005; 13: 172-8.
http://dx.doi.org/10.1109/TNSRE.2005.847369
[65] Zhang D, Maye A, Gao X, Hong B, Engel AK, Gao S. An
independent brain-computer interface using covert non-
spatial visual selective attention. J Neural Eng 2010; 7:
16010.
http://dx.doi.org/10.1088/1741-2560/7/1/016010
[66] Allison BZ, McFarland DJ, Schalk G, Zheng SD, Jackson
MM, Wolpaw JR. Towards an independent brain-computer
interface using steady state visual evoked potentials. Clin
Neurophysiol 2008; 119: 399-408.
http://dx.doi.org/10.1016/j.clinph.2007.09.121
[67] Daly JJ, Wolpaw JR. Brain-computer inter-faces in
neurological rehabilitation. Lancet Neurol 2008; 7: 1032-43.
http://dx.doi.org/10.1016/S1474-4422(08)70223-0
[68] Kübler A, Kotchoubey B, Kaiser J, Wolpaw J, Birbaumer N.
Brain-computer communication: unlocking the locked-in.
Psychol Bull 2001; 127: 358-75.
http://dx.doi.org/10.1037/0033-2909.127.3.358
[69] Mak JN, Wolpaw JR. clinical applications of brain-computer
interfaces: Current state and future prospects. IEEE Rev
Biomed Eng 2009; 2: 187-99.
http://dx.doi.org/10.1109/RBME.2009.2035356
[70] Nijboer F, Furdea A, Gunst I, Mellinger J, McFarland DJ,
Birbaumer N, et al. An auditory brain-computer interface
(BCI). J Neurosci Methods 2008; 167: 43-50.
http://dx.doi.org/10.1016/j.jneumeth.2007.02.009
[71] Allison BZ, Wolpaw EW, Wolpaw JR. Brain computer
interface systems: progress and prospects. Expert Rev Med
Devices 2007; 4: 463-74.
http://dx.doi.org/10.1586/17434440.4.4.463
[72] Birbaumer N. Brain-computer interface research: coming of
age. Clin Neurophysiol 2006; 117: 479-83.
http://dx.doi.org/10.1016/j.clinph.2005.11.002
[73] McFarland DJ, Sarnacki WA, Vaughan TM, Wolpaw JR.
Brain-computer interface (BCI) operation: signal and noise
during early training sessions. Clin Neurophysiol 2005; 116:
56-62.
http://dx.doi.org/10.1016/j.clinph.2004.07.004
[74] Pfurtscheller G, Graimann B, Neuper C. EEG-based brain-
computer interface systems and signal processing. In:
MetinAkay, eds. Encyclopedia of Biomedical Engineering.
New Jersey: John Wiley & Sons, 2006; 1156-66.
http://dx.doi.org/10.1016/S0304-3940(03)00947-9
[75] Pfurtscheller G, Müller GR, Pfurtscheller J, Gerner HJ, Rupp
R. Thought-control of functional electrical stimulation to
restore hand grasp in a patient with tetraplegia. Neurosci Lett
2003; 351: 33-6.
[76] Birch GE, Bozorgzadeh Z, Mason SG. Initial on-line
evaluations of the LF-ASD brain-computer interface with
able-bodied and spinal-cord subjects using imagined
voluntary motor potentials. IEEE Trans Neural Syst Rehabil
Eng 2002; 10: 219-24.
http://dx.doi.org/10.1109/TNSRE.2002.806839
[77] Millan J del R, Mourino J. Asynchronous BCI and local neural
classifiers: an overview of the adaptive brain interface
Machado et al.
project. IEEE Trans Neural Syst Rehabil Eng 2003; 11: 159-
61.
http://dx.doi.org/10.1109/TNSRE.2003.814435
[78] Scherer R, Müller GR, Neuper C, Graimann B, Pfurtscheller
G. An asynchronously controlled EEG-based virtual
keyboard: Improvement of the spelling rate. IEEE Trans
Biomed Eng 2004; 51: 979-84.
http://dx.doi.org/10.1109/TBME.2004.827062
[79] Pfurtscheller G, Leeb R, Keinrath C, Fried-man D, Neuper C,
Guger C, et al. Walking from thought. Brain Res 2006; 1071:
145-52.
http://dx.doi.org/10.1016/j.brainres.2005.11.083
[80] Allison BZ, Brunner C, Kaiser V, Müller-Putz GR, Neuper C,
Pfurtscheller G. Toward a hybrid brain-computer interface
based on imagined movement and visual attention. J Neural
Eng 2010; 7: 26007.
http://dx.doi.org/10.1088/1741-2560/7/2/026007
[81] Pfurtscheller G, Allison BZ, Brunner C, Bauernfeind G, Solis-
Escalante T, Scherer R, et al. The hybrid BCI. Front Neurosci
2010; 4: 42.
[82] Vaughan TM, McFarland DJ, Schalk G, Sarnacki WA,
Krusienski DJ, Sellers EW, et al. The Wadsworth BCI
Research and Development Program: at home with BCI.
IEEE Trans Neural Syst Rehabil Eng 2006; 14: 229-33.
http://dx.doi.org/10.1109/TNSRE.2006.875577
[83] Schalk G, Miller KJ, Anderson NR, Wilson JA, Smyth MD,
Ojemann JG, et al. Two-dimensional movement control using
electro-corticographic signals in humans. J Neural Eng 2008;
5: 75-84.
http://dx.doi.org/10.1088/1741-2560/5/1/008
[84] Muller KR, Tangermann M, Dornhege G, Krauledat M, Curio
G, Blankertz B. Machine learning for real-time single-trial
EEG-analysis: From brain-computer interfacing to mental
state monitoring. J Neurosci Methods 2008; 167: 82-90.
http://dx.doi.org/10.1016/j.jneumeth.2007.09.022
[85] Kübler A, Neumann N, Kaiser J, Kotchoubey B, Hinterberger
T, Birbaumer N. Brain-computer communication: Self-
regulation of slow cortical potentials for verbal
communication. Arch Phys Med Rehabil 2001b; 82: 1533-9.
http://dx.doi.org/10.1053/apmr.2001.26621
[86] Hinterberger T, Nijboer F, Kübler A, Matuz T, Furdea A,
Mochty U, et al. Brain- computer interfaces for
communication in paralysis: A clinical experimental
approach. In: Dornhege G, Millán J del R, Hinterberger T,
McFarland DJ, Müller Klaus-Robert, Eds. Toward brain-
computer interfacing. Cambridge, MA: The MIT Press, 2007;
pp. 43-64.
[87] Ang KK, Guan C, Chua KS, Ang BT, KuahC, Wang C, et al.
A clinical study of motor imagery-based brain-computer
interface for upper limb robotic rehabilitation. Conf Proc IEEE
Eng Med Biol Soc 2009; 2009: 5981-4.
[88] Broetz D, Braun C, Weber C, Soekadar SR, Caria A,
Birbaumer N. Combination of brain computer interface
training and goal-directed physical therapy in chronic stroke:
a case report. Neurorehabil Neural Repair 2010; 24: 674-9.
http://dx.doi.org/10.1177/1545968310368683
[89] Kotchoubey B, Strehl U, Uhlmann C, Holzapfel S, König M,
FröscherW, et al. Modification of slow cortical potentials in
patients with refractory epilepsy: a controlled outcome study.
Epilepsia 2001; 42: 406-16.
http://dx.doi.org/10.1046/j.1528-1157.2001.22200.x
[90] Rockstroh B, Elbert T, Birbaumer N, Wolf P, Düchting-Röth
A, Reker M, et al. Cortical self-regulation in patients with
epilepsies. Epilepsy Res 1993; 14: 63-72.
http://dx.doi.org/10.1016/0920-1211(93)90075-I
[91] Fuchs T, Birbaumer N, Lutzenberger W, Gruzelier JH, Kaiser
J. Neurofeedback training for attention-deficit/hyperactivity
Progress and Prospects in EEG-Based Brain-Computer Interface Journal of Rehabilitation Robotics, 2013, Vol. 1, No. 1 41

disorder in children: a comparison with methylphenidate.
Appl Psychophysiol Bio-feedback 2003; 28: 1-12.
http://dx.doi.org/10.1023/A:1022353731579
[92] Strehl U, Leins U, Goth G, Klinger C, Hinterberger T,
Birbaumer N. Self-regulation of slow cortical potentials – a
new treatment for children with ADHD. Pediatrics 2006; 118:
1530-40.
http://dx.doi.org/10.1542/peds.2005-2478

Received on 27-02-2013 Accepted on 22-04-2013 Published on 30-06-2013

DOI: http://dx.doi.org/10.12970/2308-8354.2013.01.01.4

© 2013 Machado et al.; Licensee Synergy Publishers.

This is an open access article licensed under the terms of the Creative Commons Attribution Non-Commercial License
(http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted, non-commercial use, distribution and reproduction in
any medium, provided the work is properly cited.