BCCI - A BIDIRECTIONAL CORTICAL COMMUNICATION

INTERFACE

A. Walter, M. Bensch, D. Brugger, W. Rosenstiel

Department of Computer Engineering, Eberhard-Karls-University Tuebingen, Tuebingen, Germany

M. Bogdan

Department of Computer Engineering, University of Leipzig, Leipzig, Germany

N. Birbaumer

Institute of Medical Psychology and Behavioural Neurobiology, University of Tuebingen, Tuebingen, Germany

A. Gharabaghi

Functional and Restorative Neurosurgery Unit, Department of Neurosurgery, and Neuroprosthetics Research Group

Werner Reichardt Centre for Integrative Neuroscience, University Hospital, Tuebingen, Germany

Keywords:

Brain-computer interface, ALS, Stroke, Cortical stimulation, Adaptive stimulation.

Abstract:

Therapeutic methods based on efferent signals from the patients’ brain have been studied extensively in the

ﬁeld of brain-computer interfaces and applied to paralysed and stroke patients. Invasive stimulation is used as

a therapeutic tool for patients with Parkinson disease, intractable chronic pain and other neurological diseases.

We give a short review of currently used applications for cortical stimulation for stroke patients and brain-

computer interfaces for paralyzed patients. We propose a reﬁned approach for stroke rehabilitation as well as

the extension of the use of invasive cortical stimulation to ALS patients with an experimental setup inspired

by classical conditioning to facilitate the communication with brain-computer interfaces for LIS and CLIS

patients. A closed-loop system is described with sophisticated methods for the identiﬁcation of recording and

stimulation sites, feature extraction and adaptation of stimulation algorithms to the patient in order to design a

bidirectional cortical communication interface (BCCI).

1 INTRODUCTION

Brain-computer interfaces (BCI) are a tool for com-

munication without the need of muscle use by mod-

ulation of brain waves and brain states by the user.

These modulations can be detected online or ofﬂine

by feature extraction algorithms and fed to classi-

ﬁcation or regression algorithms that try to deter-

mine the intention of the BCI user. The aim of

these BCI systems is to allow communication with

the surrounding world even if the normal muscular

communication channels are severly impaired. Pos-

sible applications can be found in the ﬁeld of neu-

roprosthetics, for example control of a wheelchair

or a robotic limb or they can be designed to en-

able patients to select letters in order to spell sen-

tences (Birbaumer et al., 1999)(Donchin et al., 2000).

There exist different paradigms for BCI, for exam-

ple slow cortical potentials, sensory-motor rhythms

or P300. A detailed description of these paradigms is

found in (Birbaumer and Cohen, 2007). The signals

are measured most commonly by electroencephalog-

raphy (EEG), but magnetoencephalography (MEG)

and blood-oxygen-level dependent functional MRI

(fMRI) BCIs are also used as non-invasive methods.

The advantage of an invasive approach is a better

signal quality. Penetrating multi-electrode arrays are

used in experimental studies, but epicortical grids are

more promising in regular clinical applications be-

cause of the good tradeoff between procedural risk

and signal quality.

Patients that can beneﬁt from the use of BCIs as a

communication tool might be suffering from diseases

like brain stem lesion, spinal cord injury or amy-

otrophic lateral sclerosis (ALS). These conditions can

lead to a locked-in syndrome (LIS) which means that

the patient is severly impaired in his ability to inter-

act with the outside world. Stroke patients on the

440

Walter A., Bensch M., Brugger D., Rosenstiel W., Bogdan M., Birbaumer N. and Gharabaghi A. (2009).

BCCI - A BIDIRECTIONAL CORTICAL COMMUNICATION INTERFACE.

In Proceedings of the International Joint Conference on Computational Intelligence, pages 440-445

DOI: 10.5220/0002318304400445

 SciTePress

other hand can use brain signals as an addition to

standard rehabilitative actions such as physiotherapy

(Buch et al., 2008). We propose in this paper a new

approach for communication with locked-in patients

as well as a therapeutic approach in stroke rehabilita-

tion using epicortical electrical stimulation. The stim-

ulation parameters will be adapted to the background

activity that is measured before stimulus onset. Back-

ground activity is assumed to profoundly change the

properties of signal transmission of neurons and thus

interfere with the activity evoked by stimulation (Des-

texhe et al., 2003). Animal experiments have shown,

that a closed-loop system for adaptation of the param-

eters is feasible, at least for single-electrode measure-

ments (Brugger et al., 2008).

The paper is structured as follows: We describe in

section 2 state of the BCI systems for ALS and stroke

patients. Section 3 contains an overviewover the ther-

apeutic application of cortical stimulation for stroke

patients. We present in section 4 our new approach

for a closed-loop system in stroke rehabilitation and

as an improvement for communication with ALS pa-

tients.

2 BRAIN-COMPUTER

INTERFACES FOR ALS AND

STROKE

2.1 ALS

Amyotrophic lateral sclerosis is an adult-onset mo-

tor neuron disease characterized by degeneration of

the ﬁrst and second motor neurons (Lakerveld et al.,

2008). Over the course of the disease, the pa-

tient gradually loses control of the muscles, devel-

ops weakness and spasticity and dies from respira-

tory failure usually within a few years, unless arti-

ﬁcially ventilated and fed. Cognitive functions are

said to be spared even in the latest stages of ALS

except for patients with frontal lobe dementia (Lak-

erveld et al., 2008). The patient might communicate

in this state by controlling devices with single mus-

cles. Therefore, a BCI can be useful to allow the

patient to carry out complex tasks, for example the

control of a web browser (Bensch et al., 2007) and to

provide a communication channel after muscular con-

trol has been lost. There has been extensive research

on the use of BCIs for ALS patients and about 75 %

of these patients was able to control a BCI (Kuebler

and Kotchoubey, 2007). However, none of these re-

sults could be transferred to patients in the completely

locked-in state (CLIS) and not a single CLIS patient

has regained communication via a BCI (Birbaumer

and Cohen, 2007). Because of this, new approaches

for BCIs have to be tested to train LIS patients in their

use, hoping that the training effects might carry over

to the CLIS state.

2.2 Stroke

Stroke is a leading cause of paralysis and disability

worldwide with several hundred thousand incidents

per year. The outcome of such an incident depends

heavily on the location and the size of the stroke

area, but is fatal in about a third of all cases during

the ﬁrst year after the stroke and leaves most of the

surivors with persisting neurological deﬁcits. A pos-

sible deﬁcit is a movement impairement on the con-

tralateral side, if the stroke affects motor areas like the

primary motor cortex M1 or the supplementary motor

area (SMA). Rehabilitative procedures for movement

impaired stroke patients consist mostly of physiother-

apy, which can lead by itself to improvements espe-

cially in lower limb functions like standing or walk-

ing. Therapy of upper limb function on the other hand

still needs to be improved (French et al., 2007).

Brain computer interfaces are a promising tool to aid

in the rehabilitation process of stroke patients. This

idea is based on the standard BCI paradigm of motor-

imagery experiments. Intuitively, the best way to im-

proverehabilitation is to couple the physiotherapyand

the movement intention of the patient in order to get a

causal relationship between the planning of the move-

ment, which should still be possible, and the sensory

feedback of the movement. One paradigm in BCI re-

search is ‘motor-imagery’tasks. The patient imagines

movements of different limbs, for example foot, hand

or tongue. This leads to a detectable event-related

desynchronization of the µ-rhythm originating in sen-

sorimotor areas (Pfurtscheller et al., 2005). In the

MEG BCI study of Buch (Buch et al., 2008), patients

were asked to imagine movements of the paralyzed

hand which were detected by the BCI system and ef-

fected the timed openings and closings of the orthosis.

The patients were able to learn to reliably control the

orthosis. This ability did not lead to functional reha-

bilitation of the affected hand.

3 CORTICAL STIMULATION

FOR STROKE

REHABILITATION

Since the 1990s, non-invasive stimulation has been

used on stroke patients for prognostic and diagnostic

BCCI - A BIDIRECTIONAL CORTICAL COMMUNICATION INTERFACE

441

Figure 1: Overview of a stimulation paradigm for stroke patients (top) and locked-in patients (bottom).

measure as well as a tool for pathophysiological re-

search. Transcranial magnetic stimulation (TMS) and

transcranial direct current stimulation (tDCS) are the

best known representatives of this group of methods.

TDCS is applied directly on the scalp with large

sponge electrodes soaked in saline. A direct cur-

rent of a few milliampere is applied to the electrodes.

Even if most of the current travels directly through the

scalp because of the low conductivity of the skull, this

method still inﬂuences brain tissue.

TMS on the other hand utilizes strong magnetic ﬁelds

typically with a coil placed directly above the area of

interest. Magnetic ﬁelds are less inﬂuenced by the

low conductivity of the skull than electrical currents

which leads to induction of currents in the underlying

brain tissue. Measureable aftereffects of this method

can last for several minutes to more than an hour, de-

pending on the stimulation protocol used (Fitzgerald

et al., 2006). It is possible to enhance or decrease

cortical excitability depending on the stimulation fre-

quency of repetitive TMS and the stimulus polarity in

case of tDCS. A review on noninvasive brain stimula-

tion can be found in (Wagner et al., 2007).

In recent years, these non-invasive techniques were

applied to stroke patients to improve the success rate

of standard physiotherapy (Hummel et al., 2008). A

disadvantage of the application of tDCS is the loss of

effect focality as the affected area is stretched out be-

tween two large electrodes. TMS is not affected by

this issue, but the aftereffects of TMS are not long-

lasting.

A possibly powerful alternative is invasive cortical

stimulation by epicortical electrode strips. A cran-

iotomy is necessary for the implantation, bearing the

risk of infections or other complications, but ﬁrst

studies with implanted electrodes show promising

results without serious complications (Levy et al.,

2008)(Brownet al., 2006). Implantabledeviceswhich

act autonomously or may be switched on and off by a

handheld wand offer the possibility of cortical stimu-

lation during physiotherapy. In 2007, Northstar Neu-

roscience conducted a multicenter Phase III clinical

trial that tested a prototype of an implanted stimula-

tor for stroke recovery with 146 patients (Levy et al.,

2008). The results were promising, but the effectsizes

were not as high as expected. The fraction of pa-

tients improving on the Upper Extremity Fugl-Meyer

Scale or the Arm Motor Ability Test did not differ sig-

niﬁcantly between the physiotherapy group and the

group that additionally received cortical stimulation.

Our goal is to improve this result with a closed-loop

stimulus control that utilizes the capacity of implanted

electrodes to be used for stimulation and recording si-

multaneously. This opens a new ﬁeld that is still un-

explored to the best of our knowledge: Invasive stim-

ulation of stroke patients for rehabilitation based on

simultaneous recorded background activity. A possi-

ble stimulation paradigm is summarised in ﬁgure 1.

4 A CLOSED-LOOP SYSTEM

FOR STROKE AND ALS

There are at least three issues that need to be ad-

dressed in order to provide optimized stimulation for

the patient (Plow et al., 2009): (1) Identiﬁcation of the

target site and electrode placement, (2) models for the

effect of the stimulation and (3) ﬁnding the optimal

stimulation pattern depending on the disease and the

patient.

IJCCI 2009 - International Joint Conference on Computational Intelligence

442

4.1 Electrode Placement

The target area will be determined preoperatively by

fMRI guided TMS. EMG measurements will be used

to track the arm movements initiated by the patient

as well as those evoked by TMS. During surgery, the

neurosurgeon will perform functional mapping of the

area of interest on the motor cortex to decide upon the

exact electrode placement. This is an improvement to

earlier studies that used only fMRI to place the elec-

trodes (Plow et al., 2009).

4.2 Modeling the Current Flow

The speciﬁc anatomy, the pathologic alterations and

the inﬂux of cortico-spinal ﬂuid change the conduc-

tivity with respect to healthy tissue due to the creation

of new shunting routes for the currents, which can

distort the electric ﬁeld produced by the stimulation

(Wagner et al., 2006). Thus, we need to ensure that

the desired brain regions are targeted by the stimula-

tion (Plow et al., 2009). We will conduct modeling

studies of the stimulation that show volume conduc-

tion effects by a ﬁnite element model (FEM). Stan-

dard models range in complexity from simple spher-

ical head models to complex realistic head models

based on MRI measurements. The different tissue

classes are modeled with conductivities based on val-

ues measured during experiments. Models of the elec-

trodes will be placed in the vicinity of the lesion to

investigate the effect of stimulation on the tissue. The

integration of areas affected by stroke based on MR

measurements of the patients into the model will lead

to further improvements of the accuracy of predic-

tions on the current spread by volume conduction.

4.3 Patient-speciﬁc Stimulation

4.3.1 ALS

Hypotheses, why ALS patients are not able to com-

municate with BCIs in the completely locked-in state

include: (1) difﬁculty in performing the task because

of cognitivedeﬁcits or lack of alertness (2) inability to

modulate the cortical rhythms due to degeneration or

missing feedback (3) unwillingness to cooperate (Hill

et al., 2005). Especially because of (1), we need to

ensure that the ALS patients are in a suitable cogni-

tive state.

Cognitive deﬁcits, for example frontotemporal de-

mentia have been reported in ALS, so we will use

cognitive tests to make sure that the patient is able

to understand the instructions and operate the BCI.

The alertness of the patient should be as high as pos-

sible during the experiments. Late-stage ALS patients

can not report on their current level of alertness and

thus, the researcher has to ensure that the patient is not

drowsy or sleeping during the experiment. There are

some peripheral measures that can be used as an indi-

cator for alertness, for example the heart rate. A spec-

tral analysis of the EEG or ECoG can also be helpful,

as lack of alertness can be correlated to changes in the

alpha and beta frequencies (Jung et al., 1997).

We include a third option: the analysis of connectivity

patterns in the ECoG during BCI experiments based

on phase synchronization and causality measures. It

can be assumed, that the BCI performance of patients

(which can be measured by the classiﬁcation error

of the BCI system) is to a certain degree related to

their state of alertness. As an extreme example: One

can expect, that the performance of a very drowsy or

even sleeping person will be found to be somewhere

around chance level.

With this in mind, an alertness classiﬁcation to

identify the relevant cortical activity and/or spectral

patterns should be possible if we link the signals

recorded before each trial to the BCI performance of

the trial. We will use the activity or spectral patterns

associated with particularly good and bad BCI per-

formance to assess the alertness of the patient before

an experimental session takes place. This evaluation

will help to decide whether an experiment should be

conducted, or if additional methods should be used to

stimulate the alertness of the patient.

Degeneration of nervous tissue and cortical connec-

tions in ALS may stem from two processes: On the

one hand direct pathological degeneration that pre-

vents the patients from controlled use of their mus-

cles and on the other hand the missing feedback due

to the underused muscles. Our theory here is: if the

patient is not able to affect his environment, the miss-

ing feedback leads to functional deterioration. As a

result, the patient is unable to reactivate these connec-

tions which has a negative effect not only on muscle

control, but might also impair his ability to imagine

muscle movement for a BCI based on µ rhythm mod-

ulation (Pfurtscheller et al., 2005).

P300 BCIs and slow cortical potential BCIs rely on

visual stimulus presentation and visual feedback re-

spectively. The inability of ALS patients in CLIS to

operate these types of BCI may stem from the fact

that eye focussing is also affected by the disease (Bir-

baumer and Cohen, 2007). Thus, the visual system

is not a good communication channel for ALS pa-

tients. There exist auditory P300 systems, but their

efﬁciency for CLIS patients has not been tested yet.

Our hypothesis here is, that feedback applied by cor-

BCCI - A BIDIRECTIONAL CORTICAL COMMUNICATION INTERFACE

443

tical stimulation to the patient’s brain can also be used

to counter the effects of neural degeneration.

Depending on the scope of control the stimulation has

over the evoked activity in terms of amplitudes and

duration, we may be able to imprint cortical activity

that was measured in earlier sessions of the same pa-

tient. This is a potentially interesting approach for

the improvement of communication with ALS pa-

tients: When they have learned to operate a BCI in

the locked-in state, we will store the relevant activity

patterns for later use. If the BCI performance drops

during the transition from LIS to CLIS, imprinting of

the patterns recorded earlier can be considered as a

form of classical conditioning and may slow down the

functional deterioration.

If we are not able to perform the imprinting with suf-

ﬁcient precision, a simpler approach also inspired by

classical conditioning is still feasible. We will gen-

erate two stimulation patterns for each patient, one

representing ’No’, the other one ’Yes’. True and false

statements can then be used in conjunction with the

respective stimulation pattern to condition the patient

to involuntarily modulate brain states in the absence

of stimulation. This provides the patient with a bi-

nary communication device.

While the neurosurgical procedure, the presence of

the grid in the skull and the electrical stimulation

imposes a risk on the patient, the inability of cur-

rent BCIs to establish a communication channel with

CLIS patients enforces the test of alternative ap-

proaches.

4.3.2 Stroke

The experimental setup for stroke patients will in-

clude the hand orthosis used in (Buch et al., 2008)

and a robotic arm to move the arm of the patient. The

patient will imagine movementsof the paralyzed hand

that are identiﬁed by a BCI system. If a hand move-

ment imagination is detected, the orthosis opens or

closes the hand of the patient, while elbow or shoul-

der movement imaginations trigger movements of the

robotic arm. Additionally, electrical cortical stimula-

tion is applied to the ipsilesional cortex. We will use

it to enhance cortical excitability near the lesion in

the ﬁrst patients to promote regeneration. As a reﬁne-

ment to existing stroke studies like (Levy et al., 2008),

the stimulation electrodes and parameters will depend

on results of functional mapping, volume conduction

modeling and the measured background activity.

If most of the brain tissue associated with upper limb

function is destroyed by the stroke, cortical stimula-

tion can be used to condition unaffected ipsilesional

motor cortex areas to upper limb movements. The

feasibility of this was shown by (Jackson et al., 2006)

with intracortical microstimulation in macaque mon-

keys.

4.3.3 Stimulation Parameters

Stimulation for cortical mapping or motor cortex

stimulation for patients with intractable chronic pain

consists of trains of short pulses, applied with fre-

quencies around 50-100 Hz (Brown, 2001)(Franzini

et al., 2003), which was found in animal studies to

be a frequency range that enhances local cortical ex-

citability (Teskey et al., 2003). Frequencies of 50 or

100 Hz were also used in the Northstar Neuroscience

multicenter stroke study.

As the space of all possible stimulation patterns is

very high, especially if one takes into account stim-

ulating on multiple electrodes at the same time with

a possibly free form stimulus, systematically testing

all patterns on patients is out of the question. Thus,

we will start with standard stimulation paradigms and

record the effects of the stimulation simultaneously

with the ECoG electrodes in order to analyse them

and improve the stimulation in later experiments.

We use these ﬁrst experiments to investigate the pa-

rameters that establish a functional relationship be-

tween the recorded data before the stimulation, the

delivered stimulus and the stimulus evoked potentials.

This connection between stimulation parameters and

the cortical evokedpotentials will then be used to con-

struct a closed-loop system (Brugger et al., 2008). It

will enable us to adapt the stimulation parameters to

the ongoing activity and will allow prespeciﬁed target

activities to be evoked by adaptive stimulation.

5 CONCLUSIONS

We propose here an extension to classical brain-

computer interfaces that enhances the afferent path-

way to the patient’s brain using electrical stimulation

by epidural electrodes controlled by a closed-loop

system for recording, feature extraction and stimula-

tion. It is summarised in ﬁgure 1. We believe, that

our sophisticated approach for the placement of the

epidural electrodes, the stimulation patterns used and,

in case of stroke patients, the physiotherapy will lead

to therapeutic improvements for stroke patients com-

pared to standard physiotherapy and ﬁrst studies on

applications of cortical stimulation for patients. Cur-

rent noninvasive BCI methods were not able to estab-

lish a communication channel with CLIS patients up

to this date. Because of this are ALS patients a sec-

ond target group of our bidirectional cortical commu-

nication interface. They might beneﬁt by an improved

IJCCI 2009 - International Joint Conference on Computational Intelligence

444

ability to control brain-computer interfaces even in

the completely locked-in state due to the direct inter-

face for feedback to the patient’s brain which does not

rely on possibly impaired sensory systems.

ACKNOWLEDGEMENTS

This work was supported by German Research Foun-

dation (DFG GH 94/2-1), the Federal Ministry of Ed-

ucation and Research (BMBF Bernstein 01GQ0761,

BMBF 16SV3783) and the European Union (ERC

227632).

REFERENCES

Bensch, M., Karim, A., Mellinger, J., Hinterberger, T.,

Tangermann, M., Bogdan, M., Rosenstiel, W., and

Birbaumer, N. (2007). Nessi: An EEG controlled

web browser for severely paralyzed patients. Compu-

tational Intelligence and Neuroscience, 2007:71863.

Birbaumer, N. and Cohen, L. (2007). Brain-computer inter-

faces: communication and restoration of movement in

paralysis. J. Physiol., 579:621–636.

Birbaumer, N., Ghanayim, N., Hinterberger, T., Iversen, I.,

Kotchoubey, B., Kbler, A., Perelmouter, J., Taub, E.,

and Flor, H. (1999). A spelling device for the paral-

ysed. Nature, 398(6725):297–298.

Brown, J., Lutsep, H., Weinand, M., and Cramer, S. (2006).

Motor Cortex Stimulation for the Enhancement of

Recovery from Stroke: A Prospective, Multicenter

Safety Study. Neurosurgery, 58:464–473.

Brown, J. A. (2001). Motor Cortex Stimulation. Neurosurg

Focus, 11(3):A5.

Brugger, D., Butovas, S., Bogdan, M., Schwarz, C., and

Rosenstiel, W. (2008). Direct and inverse solution for

a stimulus adaptation problem using SVR. In Pro-

ceedings of the 16th European Symposium on Artiﬁ-

cial Neural Networks. d-side Publishing.

Buch, E., Weber, C., Cohen, L., Braun, C., Dimyan,

M., Ard, T., Mellinger, J., Caria, A., Soekadar, S.,

Fourkas, A., and Birbaumer, N. (2008). Think to

Move: a Neuromagnetic Brain-Computer Interface

(BCI) System for Chronic Stroke. Stroke, 39:910–917.

Destexhe, A., Rudolph, M., and Pare, D. (2003). The high-

conductance state of neocortical neurons in vivo. Nat

Rev Neurosci, 4:739–751.

Donchin, E., Spencer, K., and Wijesinghe, R. (2000). The

Mental Prosthesis: Assessing the Speed of a P300-

Based BrainComputer Interface. IEEE TRans on Re-

hab Eng, 8:174–179.

Fitzgerald, P., Fountain, S., and Daskalakis, Z. (2006). A

comprehensive review of the effects of rTMS on mo-

tor cortical excitability and inhibition. Clinical Neu-

rophysiology, 117:2584–2596.

Franzini, A., Ferroli, P., Donges, I., Marras, C., and Broggi,

G. (2003). Chronic motor cortex stimulation for

movement disorders: A promising perspective. Neu-

rological Research, 25:123–126.

French, B., Thomas, L., Leathley, M., Sutton, C., McAdam,

J., Forster, A., Langhorne, P., Price, C., Walker, A.,

and Watkins, C. (2007). Repetitive task training for

improving functional ability after stroke. Cochrane

Database of Systematic Review, 4:CD006073.

Hill, N., Lal, T., Schroeder, M., Hinterberger, T., Wil-

helm, B., Nijboer, F., Mochty, U., Widman, G., Elger,

C., Schoelkopf, B., Kuebler, A., and Birbaumer, N.

(2005). Classifying EEG and ECoG Signals without

Subject Training for Fast BCI Implementation: Com-

parison of Non-Paralysed and Completely Paralysed

Subjects. IEEE Transactions on Neural Systems and

Rehabilitation Engineering, 14:183–186.

Hummel, F., Celnik, P., Pascual-Leone, A., Fregni, F., By-

blow, W., Bueteﬁsch, C., Rothwell, J., Cohen, L., and

Gerloff, C. (2008). Controversy: Noninvasive and in-

vasive cortical stimulation show efﬁcacy in treating

stroke patients. Brain Stimulation, 1:370–382.

Jackson, A., Mavoori, J., and Fetz, E. (2006). Long-term

motor cortex plasticity induced by an electronic neural

implant. Nature, 444:56–60.

Jung, T., Makeig, S., Stensmo, M., and Seinovsky, T.

(1997). Estimating Alertness from the EEG Power

Spectrum. IEEE Trans Biomed Eng, 44:60–69.

Kuebler, A. and Kotchoubey, B. (2007). Brain-computer

interfaces in the continuum of consciousness. Current

Opinion in Neurology, 20:643–649.

Lakerveld, J., Kotchoubey, B., and Kuebler, A. (2008).

Cognitive function in patients with late stage amy-

otrophic lateral sclerosis. J. Neurol. Neurosurg. Psy-

chiatry, 79:25–29.

Levy, R., Ruland, S., Weinand, M., Lowry, D., Dafer, R.,

and Bakay, R. (2008). Cortical stimulation for the

rehabilitation of patients with hemiparetic stroke: a

multicenter feasibility study of safety and efﬁcacy. J

Neurosurg, 108:707–714.

Pfurtscheller, G., Brunner, C., Schloegl, A., and da Silva,

F. L. (2005). Mu rhythm (de)synchronization and

EEG single-trial classiﬁcation of different motor im-

agery task. NeuroImage, 31:153–159.

Plow, E., Carey, J., Nudo, R., and Pascual-Leone, A. (2009).

Invasive Cortical Stimulation to Promote Recovery of

Function After Stroke. A Critical Appraisal. Stroke,

40:1926–1931.

Teskey, G., Flynn, C., Goertzen, C., Monﬁls, M., and

Young, N. (2003). Cortical stimulation improves

skilled forelimb use following a focal ischemic infarct

in the rat. Neurol Res, 25:794–800.

Wagner, T., Fregni, F., Eden, U., Ramos-Estebanez, C.,

Grodzinsky, A., Zahn, M., and Pascual-Leone, A.

(2006). Transcranial magnetic stimulation and stroke:

A computer-based human model study. NeuroImage,

30:857–870.

Wagner, T., Valero-Cabre, A., and Pascual-Leone, A.

(2007). Noninvasive human brain stimulation. Annu

Rev Biomed Eng, 9:527–565.

BCCI - A BIDIRECTIONAL CORTICAL COMMUNICATION INTERFACE

445