Professional Documents
Culture Documents
26
Noninvasive Brain–Computer Interfaces
Gerwin Schalk1,2,3, Brendan Z. Allison1,4
1National Center for Adaptive Neurotechnologies, Albany, NY, United States; 2Albany Medical College,
Albany, NY, United States; 3State University of New York at Albany, Albany, NY, United States;
4University of California San Diego, La Jolla, CA, United States
O U T L I N E
INTRODUCTION
BCIs may replace lost functions, such as speaking
or moving. They may restore the ability to control the
Overview of This Chapter
body, such as by stimulating nerves or muscles that
Brain–computer interfaces (BCIs) measure brain move the hand. BCIs have also been used to improve
activity, extract features from that activity, and convert functions, such as training users to improve the
those features into outputs that replace, restore, enhance, remaining function of damaged pathways required to
supplement, or improve human functions. grasp. BCIs can also enhance function, like warning a
AMPLITUDE (µV)
0.6
2 10 0.4
0.2 r
r2
0.2
0.0 0 0.0
10 cm 0 20 40 60 0 20 40 60
FREQUENCY (Hz) FREQUENCY (Hz)
FIGURE 26.1 (A and B) The changes in mu activity centered around 12 Hz for (A) actual and (B) imagined right-hand movements. The colors
reflect the proportion of the signal variance accounted for by the task. These two images show that imagined movements produce changes that
are less pronounced than those resulting from actual movements, but show a similar topographical distribution. (C) EEG power over site C3 for a
different subject who rested (dashed line) or performed right-hand movement (solid line). The mu activity at about 12 Hz and its harmonic around
24 Hz are both greatly reduced by movement. (D) The resulting r2 correlations for rest versus movement. This image also shows that movement
primarily affects power in the mu frequency bands and its harmonics. From Schalk, G., McFarland, D.J., Hinterberger, T., Birbaumer, N., Wolpaw, J.R.,
2004. BCI2000: a general-purpose brain–computer interface (BCI) system. IEEE Trans. Biomed. Eng. 51 (6), 1034–1043.
sleepy driver to wake up. Finally, a BCI might supple- Different types of features can be detected in the
ment the body’s natural outputs, such as through a EEG and may serve as the basis for BCIs. One of the
third hand. most important of these features is oscillatory activ-
Different techniques are used to measure brain activ- ity in different frequency bands: delta (less than 4 Hz),
ity for BCIs. Most BCIs have used electrical signals that theta (4–8 Hz), alpha (8–12 Hz), beta (18–25 Hz), and
are detected using electrodes placed invasively within gamma (greater than 30 Hz). While the origin of oscil-
or on the surface of the cortex, or noninvasively on the latory activity is still debated, oscillations probably
surface of the scalp [electroencephalography (EEG)]. reflect interactions between the cortex and the thalamus
Some BCIs have been based on metabolic activity that or other subcortical structures. Delta activity is most
is measured noninvasively, such as through functional prominent during deep sleep when high-amplitude
magnetic resonance imaging (fMRI). delta waves can be prevalent over many areas. Theta
This chapter is focused on providing an overview of activity is prevalent during light sleep and meditation.
noninvasive BCIs. After a brief review of the relevant Alpha activity increases over occipital areas when peo-
aspects of EEG and fMRI, each of the subsequent sec- ple rest with their eyes closed and during light sleep,
tions is dedicated to one of the four different purposes and (along with theta and beta) may be used in BCIs to
that a BCI may serve and that have been realized as of indicate workload or concentration. The phenomenon
this writing. of “alpha blocking” refers to the decrease in alpha activ-
ity that occurs when a person is asked to open the eyes
and perform a complex task. Because this is one of the
Electroencephalography
most obvious changes in the EEG that people can eas-
EEG sensors detect the coordinated activity of ily produce, users are often asked to alternate between
large groups of neurons—the electrical signature of eyes-closed relaxation and eyes-open concentration to
individual or only a few neurons is not detectable by confirm that their EEG system is working properly.
electrodes outside the skull. EEG sensors are usually The changes in EEG activity during sleep are driven
placed in an electrode cap that is designed to position largely by activity in the pons, thalamus, and occipital
the electrodes over specific brain regions. Some work regions. Activity in the same alpha frequency range, but
has presented EEG electrodes in headbands, head- detected over sensorimotor instead of visual areas, is
phones, glasses, or other less obtrusive headwear. For called the mu rhythm. The mu rhythm is modulated by
many years, EEG electrodes were usually composed expected, actual, observed, or imagined motor move-
of silver/silver chloride rings that were housed in a ments or associated sensations. These changes in mu
plastic disk. Electrode gel was needed to establish an activity have been called event-related (de-)synchroni-
electrical connection between the scalp’s surface and zation or ERD/S (see Fig. 26.1), and have been widely
each electrode. Work has validated dry electrodes that used in BCIs.
eliminate the time and inconvenience of gel (Guger Beta and gamma activity is most apparent during con-
et al., 2012; Fridman et al., 2016), but to what extent centration and can also include harmonics of mu activ-
dry electrodes provide stable EEG, in particular in ity (Pfurtscheller, 1981; Pfurtscheller et al., 1997). These
uncontrolled environments and when used by nonex- frequency bands have been used in BCIs to detect con-
perts, is still unclear. centration or information overload. Both bands are often
AMPLITUDE (µV)
0.45 0.6
R2
0.3 0.4
0.15 0.2
0 15.00 Hz 18.00 Hz
0
0 15 30 45 0 15 30 45
(C) (D)
0.6
AMPLITUDE (µV)
0.45 0.6
30.00 Hz 45.00 Hz
R2
0.3 0.4
0.6
0.15 0.2
0.4
0 0 0.2
0 15 30 45 0 15 30 45
Frequency (Hz) Frequency (Hz) 0
FIGURE 26.3 Steady-state visual evoked potential (SSVEP) activity elicited during selective attention to two oscillating checkerboards, each of
which oscillated at 6 or 15 Hz. (A and C) Spectral power for one subject over site O1 (A) or O2 (C). The solid and dotted lines show activity elicited
while the subject focused on the 15- or 6-Hz checkerboard, respectively. (B and D) The r2 values that reflect the correlation between different fre-
quencies and the instruction to focus on either target stimulus. (E) A topographic map of these differences. It is shown that selective attention to a
flickering stimulus increases power at the eliciting frequency and, to a lesser extent, the harmonics of that frequency. The SSVEP activity is much
more pronounced over occipital areas than over other sites. From Allison, B.Z., McFarland, D.J., Schalk, G., Zheng, S.D., Jackson, M.M., Wolpaw, J.R.,
2008. Towards an independent brain–computer interface using steady state visual evoked potentials. Clin. Neurophysiol. 119 (2), 399–408.
all healthy adults (Guger et al., 2012), but have not Cursor movement has been extended to a variety of
been well explored with patients (but see Lim et al. tasks with noninvasive BCIs, including web browsing
2013; Zoltan et al. 2016). (Long et al., 2015; Mugler et al., 2010; Karim et al., 2006)
One of the most prominent BCI research directions and gaming/virtual navigation (Scherer et al., 2012;
in the late 20th century relied on slow cortical poten- Coyle et al., 2011). BCI-based control of smart homes
tials (SCPs). These are very slow drifts in the EEG that can also implement virtual navigation through a home
patients can learn to increase or decrease over months environment (Edlinger et al., 2011; Carabalona et al.,
of training, prominent over central sites. Patients 2012).
with no residual movement learned to modulate their Another way that BCIs can replace lost functions is
SCPs to move a cursor to iteratively select letters or through providing a mechanism for creative expres-
letter groups (Birbaumer et al., 1999). SCPs have not sion. BCIs have been used to compose music based on
been widely used in BCIs for several years because EEG measures of emotion (Makeig et al., 2011). The
of the long training time and low communication Brain Painting system allows users to create paintings
bandwidth. on a monitor through motor imagery or P300 activity
BCIs for spelling based on motor imagery gained (Halder et al., 2009; Muenssinger et al., 2010). Several
attention after work showed that patients with ALS ALS patients have posted their paintings online, and
can use motor imagery to control a BCI (Kuebler et al., reported significant enjoyment using the system.
2005). Several people, including a patient with tetraple-
gia, were able to use motor imagery to direct a cursor up Worn Robotic Devices
or down toward different letters or letter groups on the BCIs have been validated for control of wearable
right side of a monitor while the cursor moved steadily robotic devices such as orthoses, prostheses, and exo-
from left to right (Vaughan et al., 2006). In the Hex-O- skeletons. In Pfurtscheller et al. (2010) and Ortner
Spell approach (Blankertz et al., 2006), the user views a et al. (2011), subjects used SSVEP activity to control
monitor with a hexagon surrounded by six other hexa- a hand orthosis. In addition, in Pfurtscheller et al.
gons. The central hexagon contains an arrow, while the (2010), the system also allowed subjects to use mu
other hexagons each contain six letters or other charac- activity to activate or deactivate LEDs generating the
ters. At the start of each trial, the arrow begins moving in SSVEP. This hybrid approach allowed users to reduce
a clockwise direction. When the arrow points to a hexa- the annoyance caused by flickering stimuli. Related
gon containing the desired group of characters, the user work with BCIs to control functional electrical stimu-
can perform motor imagery (such as left hand grasping) lation, prostheses, and exoskeletons shows potential
to make the arrow longer until it reaches the desired to both replace natural mobility and facilitate therapy
hexagon. Next, the arrow returns to its starting point, (Angulo-Sherman et al., 2016; Rohm et al., 2014; Rupp,
while the other six hexagons’ contents each change to 2014).
one of the six characters that the user just chose. The
arrow begins moving again, and the user can choose one Mobile Robotic Devices
of the six characters in the same fashion. Thus, Hex-O- Bell et al. (2008) demonstrated a P300-based BCI
Spell provides an intuitive two-level spelling interface, system that presented either four or six images that
with clear trial timing and goals, based on simple binary each corresponded to robot control commands. Data
motor control (Severens et al., 2014; Muller et al., 2008; were recorded from 32 EEG channels while nine
Rohani et al., 2012). healthy subjects silently counted each time a target
image flashed. The overall accuracy across subjects
was 98.4%, yielding up to 24 bits/min. Each command
Control Functions could direct a mobile robot equipped with a camera
to perform complex actions, relying on the robot’s
Computer Functions software to perform the low-level actions needed to
The first noninvasive BCI publication described navigate around a room, get a glass, or perform other
an SSVEP-based BCI in which the user could direct tasks. In Galan et al. (2008), two healthy subjects each
a cursor up, left, down, or right by focusing on one participated in five sessions with 10 trials each, during
of four oscillating boxes on a monitor (Vidal, 1973). which they used a 64-channel EEG system to drive real
Several groups have described noninvasive BCIs for and simulated wheelchairs along a predefined path.
one-, two-, or three-dimensional cursor control (Li The BCI allowed three commands based on mental
et al., 2010; McFarland et al., 2010; Wolpaw et al., 2003; tasks (left-hand imagery, turn left; rest, forward; word
Muller et al., 2008; Coyle et al., 2011; Scherer et al., association, turn right). The two subjects attained
2012; Long et al., 2015). 100% and 80% overall accuracy.
Introduction Orthoses
BCIs may be used to restore a patient’s ability to con- Orthoses are external, noninvasive devices that are
trol his or her body. This category of BCIs aims solely to attached to the body to facilitate movement in vari-
help persons with disabilities. Unlike BCIs that replace ous ways. Simple orthoses include plastic braces that
function, which control external devices (such as a can be strapped to the foot and ankle that help restore
robotic arm), BCIs that restore function eventually move some movement to persons with foot or ankle injuries.
the body’s own limbs. The goal is to bypass damaged Simple orthoses like these have no moving parts, no
pathways that connect to functioning effectors, such as degrees of freedom, and no need for a control mecha-
the patient’s arms and hands. This technology could nism. More complex orthoses may entail mechanical
benefit people with stroke, brain injury, spinal cord components that are designed to move the foot, arm,
injury, and other conditions that damage the brain or shoulder, or other body part. These systems often have
spinal cord. more than 1 degree of freedom, and need some mecha-
A substantial volume of work has focused on restor- nism to control their operation. In principle, BCIs are
ing function to the arms and hands. This is a prevalent an appealing control mechanism, as these users already
need for many patients and is a relatively safe research have limited mobility and thus reduced options to
direction (restoring function to lower limbs adds the risk control devices. With complex orthoses, BCIs may be
of falling), and stimulating nerves or muscles in the arm used in combination with shared control so that users
to initiate hand grasping may be simpler than producing can simply imagine performing the desired movement
the extremely intricate and coordinated muscle activa- (such as walking) and the details of the timing, loca-
tions necessary for walking or speaking. The following tion, intensity, and duration of muscle stimulation are
section gives an overview of devices that have been used managed by the BCI system (Ortner et al., 2011; Rohm
to produce limb movements. et al., 2013; Millan et al., 2010).
FIGURE 26.5 (A) A patient using an electrode cap to detect movement imagery that causes a noninvasive (transcutaneous) functional electri-
cal stimulator (FES) system to activate and trigger grasp function. (B) A different patient who instead uses an invasively implanted FES to control
grasp. From Muller-Putz, G.R., Scherer, R., Pfurtscheller, G., Rupp, R., 2006. Brain–computer interfaces for control of neuroprostheses: from synchronous to
asynchronous mode of operation. Biomed. Tech. (Berl.) 51 (2), 57–63.
FIGURE 26.6 The left (A, top) shows an orthosis that can be opened or closed by moving along the axes shown by white arrows via steady-
state visual evoked potential activity elicited by either of the black LEDs (B, left middle). The remaining three images show the orthosis moving
throughout three stages of opening and closing. The right shows a healthy volunteer using the orthosis to move a bottle with a white foam shield.
From Ortner, R., Allison, B.Z., Korisek, G., Gaggl, H., Pfurtscheller, G., 2011. An SSVEP BCI to control a hand orthosis for persons with tetraplegia. IEEE
Trans. Neural Syst. Rehabil. Eng. 19 (1), 1–5.
Fig. 26.5 presents examples of two patients using dif- Fig. 26.6). Six volunteers attained good accuracy in an
ferent types of FES systems with BCIs to restore grasp asynchronous control paradigm, although the system
control. produced a high rate of false positives, and some par-
Many groups have explored other issues relating to ticipants did not like the flickering lights (Ortner et al.,
functional electrical stimulation and upper limb move- 2011). Related work addressed these problems with a
ment. Other early work (Pfurtscheller et al., 2002) com- hybrid motor imagery–SSVEP system for orthosis con-
pared the efficacy of EEG- to EMG-based control of an trol. This BCI system allowed subjects to use motor
orthosis. In one study exploring long-term use, a single imagery as a “brain switch” to switch the LEDs on or
C4 spinal cord injury (SCI) patient used EEG-based off, thus enabling users to eliminate both problems.
motor imagery measures to control both an FES and an Four of six healthy participants attained good accu-
orthosis to restore hand function. After a year of train- racy (Pfurtscheller et al., 2010). Ramos-Murguialday
ing, his motor imagery accuracy remained at only 70%, et al. (2012) found that proprioceptive feedback led to
but he still found the device useful (Rohm et al., 2013). improved motor imagery control of a robotic exoskel-
For patients who cannot attain good accuracy with eton (compared to no feedback).
motor imagery, SSVEP activity was validated for ortho-
sis control. Seven healthy volunteers without prior Lower Limb
training were told to focus on one of two flickering BCIs to restore control of the lower limbs often use
LEDs placed on an orthosis to either open or close it (see EEG-based imagination of walking, as this provides a
FIGURE 26.7 An EEG-based system to control a lower-limb exoskeleton being used by a healthy volunteer (left) and a spinal cord injury
patient (right). The work with the patient requires safety rails and nearby staff because of the risk of falling. From Lopez-Larraz, E., Trincado-Alonso, F.,
Rajasekaran, V., Perez-Nombela, S., Del-Ama, A.J., Aranda, J., Minguez, J., Gil-Agudo, A., Montesano, L., August 3, 2016. Control of an ambulatory exo-
skeleton with a brain–machine interface for spinal cord injury gait rehabilitation. Front. Neurosci. 10, 359. http://dx.doi.org/10.3389/fnins.2016.00359.
eCollection 2016.
FIGURE 26.8 The EEG-enhanced World of Warcraft system from Nijholt et al. (2009). The left shows how EEG-based measures of stress can
change the elf character into a bear. The right shows a person playing the game.
+
scalp projection
3 µV
H
500 msec
Average +3.9
µV
–0.1
END
OF
TRIAL 0.1s–0.2s
FIGURE 26.9 The left and middle show error-related activity from Schalk et al. (2000). The left image shows event-related potentials over site
Cz, and the middle image shows the scalp projections for 40 ms around the positive peak for each subject. The right shows error activity from
Parra et al. (2003) from 100 to 200 ms after feedback presentation. These images indicate prominent central activity 100–200 ms after feedback
presentation.
measures to adapt the system’s interaction with that user that requires time and attention. EEG measures can
(Schmorrow et al., 2006; St. John et al., 2004). Automatic reveal error potentials, such as event-related negativity
adaptation to the state of a user has been further explored (ERN), if people believe they just made a mistake. Real-
in aviation (Brookings et al., 1996; Sterman and Mann, time detection of ERN or similar EEG features could
1995). Kirchner et al. (2016) presented a system that be used to correct errors or other goals (Hoffmann and
could adapt ongoing interaction with a P300 BCI system Falkenstein, 2012; Wessel, 2012). Schalk et al. (2000)
based on EEG measures of task load. recorded activity from 64 EEG channels while four sub-
BCIs have also been used to enhance games and cre- jects used mu and beta activity to direct a cursor to the
ative applications. In Nijholt et al. (2009), healthy people word “yes” or “no,” and established the difference in
played World of Warcraft through conventional means EEG activity recorded during the trials in which the
(keyboard and mouse). The system changed the player’s subjects did or did not succeed in controlling the cursor
game character between an elf and a bear based on EEG- to the correct target. One hundred eighty milliseconds
based evaluation of stress (see Fig. 26.8). Reissland and after subjects received feedback indicating an errone-
Zander (2009) and Zander and Jatzev (2012) described ous selection, they exhibited a strong positive peak,
EEG-based systems to detect bluffing in a game environ- followed by a negative peak, which was most promi-
ment and perceived loss of control. Makeig et al. (2011) nent over Cz (Fig. 26.9, left). The authors estimated
described a BCI that can produce music in real time that, if error activity were used for error correction
based on EEG measures of user emotion. within their yes/no BCI system, the system’s informa-
tion transfer rate would improve by 0% to 21% across
the four subjects. A different group also explored error
Error Detection activity recorded from 64 EEG channels and implica-
People inevitably make mistakes while using BCIs or tions for online error correction (Parra et al., 2003).
other technologies. Errors may go undetected or uncor- Seven healthy subjects performed a button-press task
rected, or users need to perform some corrective action in a visual discrimination paradigm. ERPs also showed
0 0 0
1 1 1
µV
µV
µV
2 2 2
3 3 3
0 50 100 150 200 250 300 0 50 100 150 200 250 300 0 50 100 150 200 250 300
Time (msec) Time (msec) Time (msec)
FIGURE 26.10 Event-related potential activity elicited by images that varied in composition, content, and emotion. From Loew, A., Bradley,
M.M., Lang, P.J., 2013. Perceptual processing of natural scenes at rapid rates: effects of complexity, content, and emotional arousal. Cogn. Affect. Behav.
Neurosci. 13 (4), 860–868. Available from: http://dx.doi.org/10.3758/s13415-013-0179-1.
prominent frontocentral activity from 100 to 200 ms (RSVP), a sequence of single images is presented very
after erroneous responses only (Fig. 26.9, right), which quickly while EEG activity is used to distinguish target
could reduce errors by −6% to 49% across the seven from nontarget images. Gerson et al. (2005) showed that
subjects. nine subjects who had to press a button after each tar-
The P300 can also indicate that the user received get stimulus exhibited frontal EEG activity that devel-
feedback that was unexpected or erroneous (Polich, oped 200 ms before the button press. Pohlmeyer et al.
2004), which could be used in adaptive BCIs (Bayliss (2011) extended this direction with a real-time system
et al., 2004). Tong et al. (2016) used multiple features for to reorder an image database. EEG measures were com-
error detection while healthy people used a P300 BCI. bined with a semisupervised artificial visual system to
In addition to modifying users’ interactions with BCI- estimate users’ interest in each image during RSVP and
based communication systems, real-time error moni- change its position in the image sequence accordingly.
toring could also lead to adaptations with conventional In Loew et al. (2013), 27 participants observed 128 color
software, such as real-time error correction when a typ- images that varied in composition (figure/ground or
ist makes a mistake using a word processor (Allison, scene), content (people or none), and valence (arous-
2009). ing or neutral). The images were presented serially for
330 ms each with no delay between images while 128
EEG channels were recorded. Fig. 26.10 shows how
Sleep averaged ERPs differ across these three categories.
Millions of people in the United States alone have Results could be used to order stimuli based on richer
serious trouble sleeping, including falling asleep, information about each image, although extending
remaining asleep, and having nightmares. BCIs may be results to single-trial applications may be challenging.
used to enhance sleep in different ways. Sleep stages, Several articles have addressed single-trial RSVP algo-
including sleep onset, could be identified and classi- rithms and performance, as this application is not well
fied via EEG, EMG, electrooculogram (EOG), heart suited to averaged data (Fuhrmann Alpert et al., 2014;
rate (HR), and other signals (Beniczky et al., 2013; Manor and Geva, 2015; Lin et al., 2015; Bigdely-Shamlo
Goldstein and Chervin, 2016; Lin et al., 2008; Silverman et al., 2008; Marathe et al., 2016).
et al., 1968). These signals have been used to modulate
a user’s environment, such as relaxing music to foster
Neuromarketing
sleep onset, or influence the timing of an alarm clock.
The years since 2007 have seen considerable work on
neuromarketing, in which EEG, fMRI, magnetoenceph-
Image Recognition alography, and other physiological signals are used to
One approach to BCI could enhance the natural abil- help marketing researchers learn more about people’s
ity to convey the recognition of specific images. People decisions and reactions relating to products and services.
employed in image recognition, such as in security, This type of research may not fall under the conventional
research, or product review, often must view many definition of a BCI (because it does not provide real-time
thousands of images with the goal of detecting one or feedback to the user). At the same time, this approach
more “target” images, then pressing a button or taking could adapt to users and/or modify the ongoing inter-
other action to convey this target detection to a com- action with users, such as by presenting additional ads
puter. BCIs may improve this process by automatically that are similar to an ad that a user likes, to better iden-
determining when the user detects a target image, elim- tify which components of that ad are appealing. In a
inating the need to respond and allowing faster image typical paradigm, healthy participants view ads, movie
presentation. In the rapid serial visualization paradigm trailers, or other short clips while activity is monitored,
Introduction
approaches to improve motor control include therapy,
This section addresses BCIs that are designed to pro- several robotic assistive devices, functional electrical
duce a lasting, perhaps permanent, improvement in ner- stimulation, virtual reality, and other means.
vous system function to alleviate a particular condition. Since 2007, many researchers have presented real-
Thus, BCIs to improve function differ from the other BCI time brain imaging, sometimes with direct electrical or
approaches, which focus primarily on providing ben- magnetic brain stimulation, as a complement to existing
efits during BCI use. For example, BCIs to restore motor movement therapy tools and methods. The expectation
function may involve the same hardware and similar is that real-time measures of motor imagery could influ-
software as BCIs to improve motor function, but only the ence devices that present feedback (Fig. 26.11). Thus,
latter category focuses on training protocols that provide feedback such as FES activation, avatar movement, brain
lasting beneficial changes to the nervous system, and stimulation, and/or rewarding sounds or images could
thereby complement rehabilitation therapy. occur only when the patient imagines the correct move-
Successful development of BCIs to improve function ment, perhaps influenced further by movement inten-
would address an enormous problem: in addition to the sity or other parameters.
personal impact on the lives of patients and those who This combination of approaches is based on the prin-
care for them, movement and other disabilities can entail ciple that improving the neural communication between
many other costs. Patients may have limited or no ability two relevant groups of neurons relies on the coordinated
to work, create, or care for themselves or others. Assistive activation between them, as per Hebbian learning and
technologies, managed care facilities, rehabilitation sys- long-term potentiation. The goal of rehabilitation ther-
tems and therapies, and other ancillary costs are tremen- apy is to strengthen the pathways connecting neurons in
dous (Demaerschalk et al., 2010; Howard-Wilsher et al., the cortex (such as neurons responsible for grasping) to
2016). Therapy may require numerous sessions, spread neurons that directly trigger corresponding muscle activ-
across weeks or months, which can require being driven ity. BCIs could add several contributions to this process.
to a hospital or rehabilitation center. Thus, new devices, Real-time measures of movement imagery in the cortex
methods, or concepts to improve the efficacy and effec- can be used to automatically influence different types
tiveness of rehabilitation therapy are sorely needed. of feedback while providing helpful information to a
therapist, including compliance monitoring. Electrical or
magnetic devices can stimulate relevant cortical neurons
Improvements to Motor Function
to further increase activation when downstream neurons
Millions of people in the United States alone have dif- are active. Other neuroimaging advances could lead to
ficulty moving because of disease or injury affecting the more detailed and accurate diagnoses and treatments.
brain and/or spinal cord. Major causes include stroke, Beyond improving movement, BCIs can produce other
traumatic brain injury, tumors, and SCI. Conventional persistent changes to treat different patient groups.
Impaired Impaired
efferent efferent
signal signal
Impaired Practice of
CNS volitional CNS close-to-normal
movement movement
ly
ive
Br
ss
ai
re
n
sig
og
l
pr
na
na
m
ls
tio
ar
More normal afferent signal
or
Direct training of
ar
nc
e
ls
e
fu
cortical signal
fu
na
pr
n
e
og
sig
ct
or
io
re
m
ain
na
ss
l
ive
Br
ly
(C) Goal Normal
Restored volitional
CNS movement
Normal
efferent
signal
FIGURE 26.12 In (A), users directly train the central nervous system (CNS) to produce healthier signals, thereby leading to improved CNS
function (C). In (B), users instead attain this goal by practicing movements using a brain–computer interface assisted by a device, improving
sensory feedback and thus improving CNS function. From Daly, J.J., Wolpaw, J.R., 2008. Brain–computer interfaces in neurological rehabilitation. Lancet
Neurol. 7 (11), 103243.
Pfurtscheller and Neuper (2006) presented future stimulation) improved the learning of a finger move-
directions in BCI research, with prominent attention ment sequence, and produced significantly larger
to improving functional recovery from stroke. They muscle evoked potentials (MEPs) (Kim et al., 2006);
noted that detecting motor imagery is essential in effec- see also Chang et al. (2015).
tive rehabilitation, and addressed related issues that Some authors have continued to explore TMS for
remain topical, such as virtual reality training, BCIs motor rehabilitation. For example, Buetefisch et al. (2015)
that manage real-time feedback based on motor imag- showed that rTMS of 0.1 Hz, unlike other frequencies or
ery, first-person/kinesthetic motor imagery, and adapt- subthreshold stimulation, led to significant improve-
ing parameters to each patient. Birbaumer and Cohen ments in acceleration and MEP size. Du et al. (2016)
(2007) suggested the use of transcranial magnetic stim- found that TMS stimulation in first-time stroke patients
ulation (TMS) or transcranial direct current stimula- was more effective at 1 Hz than at 3 Hz, both of which
tion (TDCS) to strengthen Hebbian learning and thus were more effective than sham. These and other stud-
improve movement. ies using TMS to improve motor function support the
Daly and Wolpaw (2008) presented two BCI notion that simultaneous activation of both cortical and
approaches to improve function. In the first approach, associated downstream neurons increases motor learn-
users train to produce different patterns of brain activ- ing. Furthermore, additional work to explore optimal
ity through feedback. The second approach relies on stimulation frequencies and other parameters could lead
repeated use of BCI to control a movement device to to greater improvement.
encourage activity-dependent plasticity through sensory Mrachacz-Kersting et al. (2011) found that MRCPs can
feedback (Fig. 26.12). be used to trigger peripheral stimulation that improves
Other early work suggested TMS as a tool to accel- excitability. Mrachacz-Kersting et al. (2016) used MRCPs
erate rehabilitation of cognitive, motor, and lan- to detect movement and thereby control feedback in
guage functions (Rossi and Rossini, 2004; Siebner a rehabilitation paradigm, and described significant
et al., 1999). The former article showed that TMS improvements in functional and neuroimaging mea-
could help patients with dystonia such as writer’s sures with only three training sessions. The authors
cramp, although improvements were only transient. noted that MRCPs can provide more precise synchrony
Other early work with 15 chronic hemiparetic stroke between movement imagery and feedback, thus improv-
patients showed that repetitive (r) TMS (unlike sham ing Hebbian learning.
Guger, C., Daban, S., Sellers, E., Holzner, C., Krausz, G., Carabalona, R., Koo, B., Lee, H.G., Nam, Y., Kang, H., Koh, C.S., Shin, H.C., Choi, S., 2015.
Gramatica, F., Edlinger, G., 2009. How many people are able to con- A hybrid NIRS-EEG system for self-paced brain computer interface
trol a P300-based brain-computer interface (BCI)? Neurosci. Lett. with online motor imagery. J. Neurosci. Methods 244, 26–32.
462 (1), 94–98. Kouijzer, M.E., van Schie, H.T., Gerrits, B.J., Buitelaar, J.K., de Moor,
Guger, C., Krausz, G., Allison, B.Z., Edlinger, G., 2012. Comparison of J.M., 2013. Is EEG-biofeedback an effective treatment in autism
dry and gel based electrodes for P300 brain-computer interfaces. spectrum disorders? A randomized controlled trial. Appl.
Front. Neurosci. 6, 60. Psychophysiol. Biofeedback 38 (1), 17–28.
Halder, S., Furdea, A., Leeb, R., Müller-Putz, G., Hosle, A., Kübler, A., Krusienski, D.J., Sellers, E.W., McFarland, D.J., Vaughan, T.M., Wolpaw,
2009. Implementation of SMR-based brain painting. In: Poster at J.R., 2008. Toward enhanced P300speller performance. J. Neurosci.
COST Neuromath Workshop. Leuven, Belgium. Methods 167 (1), 15–21.
Hoffmann, S., Falkenstein, M., 2012. Predictive information processing Kubis, N., 2016. Non-invasive brain stimulation to enhance post-stroke
in the brain: errors and response monitoring. Int. J. Psychophysiol. recovery. Front. Neural Circuits 10.
83 (2), 208–212. Kuebler, A., Nijboer, F., Mellinger, J., Vaughan, T.M., Pawelzik, H.,
Hortal, E., Planelles, D., Iáñez, E., Costa, A., Úbeda, A., Azorín, J., Schalk, G., McFarland, D.J., Birbaumer, N., Wolpaw, J.R., 2005.
2016a. Detection of gait initiation through a ERD-based brain-com- Patients with ALS can use sensorimotor rhythms to operate a brain-
puter interface. In: Advances in Neurotechnology, Electronics and computer interface. Neurology 64 (10), 1775–1777.
Informatics. Springer, pp. 141–150. Lackner, N., Unterrainer, H.F., Skliris, D., Shaheen, S., Dunitz-Scheer,
Hortal, E., Ubeda, A., Ianez, E., Azorin, J.M., Fernandez, E., 2016b. M., Wood, G., Scheer, P.J., Wallner-Liebmann, S.J., Neuper, C., July–
EEG-based detection of starting and stopping during gait cycle. Int. September, 2016. EEG neurofeedback effects in the treatment of
J. Neural Syst. 26 (7), 1650029. adolescent anorexia nervosa. Eat. Disord. 24 (4), 354–374. http://
Howard-Wilsher, S., Irvine, L., Fan, H., Shakespeare, T., Suhrcke, M., dx.doi.org/10.1080/10640266.2016.1160705. Epub March 30, 2016.
Horton, S., Poland, F., Hooper, L., Song, F., 2016. Systematic over- Lauer, R.T., Peckham, P.H., Kilgore, K.L., 1999. EEG-based control of a
view of economic evaluations of health-related rehabilitation. hand grasp neuroprosthesis. Neuroreport 10 (8), 1767–1771.
Disabil. Health J. 9 (1), 11–25. Lesenfants, D., Chatelle, C., Saab, J., Laureys, S., Noirhomme, Q.,
Huang, K.C., Huang, T.Y., Chuang, C.H., King, J.T., Wang, Y.K., Lin, 2016. Neurotechnological communication with patients with dis-
C.T., Jung, T.P., 2016. An EEG-based fatigue detection and mitiga- orders of consciousness. In: Neurotechnology and Direct Brain
tion system. Int. J. Neural Syst. 26 (4), 1650018. Communication: New Insights and Responsibilities Concerning
Hwang, H.-J., Choi, H., Kim, J.-Y., Chang, W.-D., Kim, D.-W., Kim, K., Speechless but Communicative Subjects, p. 85 (Chapter 6).
Jo, S., Im, C.-H., 2016. Toward more intuitive brain-computer inter- Li, Y., Long, J., Yu, T., Yu, Z., Wang, C., Zhang, H., Guan, C., 2010. An
facing: classification of binary covert intentions using functional EEG-based BCI system for 2-D cursor control by combining mu/
near-infrared spectroscopy. J. Biomed. Opt. 21 (9), 091303. http:// beta rhythm and P300 potential. IEEE Trans. Biomed. Eng. 57 (10),
dx.doi.org/10.1117/1.JBO.21.9.091303. 2495–2505.
Jiang, N., Gizzi, L., Mrachacz-Kersting, N., Dremstrup, K., Farina, Li, Y., Pan, J., Long, J., Yu, T., Wang, F., Yu, Z., Wu, W., 2016. Multimodal
D., 2015. A brain-computer interface for single- trial detection of BCIs: target detection, multidimensional control, and awareness
gait initiation from movement related cortical potentials. Clin. evaluation in patients with disorder of consciousness. Proc. IEEE
Neurophysiol. 126 (1), 154159. 104 (2), 332–352.
Jin, J., Allison, B.Z., Kaufmann, T., Kubler, A., Zhang, Y., Wang, X., Lim, J.-H., Hwang, H.-J., Han, C.-H., Jung, K.-Y., Im, C.-H., 2013.
Cichocki, A., 2012. The changing face of P300 BCIs: a comparison Classification of binary intentions for individuals with impaired
of stimulus changes in a P300 BCI involving faces, emotion, and oculomotor function: “eyes-closed” SSVEP-based brain-computer
movement. PLoS One 7 (11), e49688. interface (BCI). J. Neural Eng. 10 (2), 026021.
Jung, T.P., Makeig, S., Stensmo, M., Sejnowski, T.J., 1997. Estimating Lin, C.T., Chen, Y.C., Huang, T.Y., Chiu, T.T., Ko, L.W., Liang, S.F.,
alertness from the EEG power spectrum. IEEE Trans. Biomed. Eng. Hsieh, H.Y., Hsu, S.H., Duann, J.R., 2008. Development of wire-
44 (1), 60–69. less brain computer interface with embedded multitask scheduling
Karim, A.A., Hinterberger, T., Richter, J., Mellinger, J., Neumann, and its application on real-time driver’s drowsiness detection and
N., Flor, H., Kübler, A., Birbaumer, N., 2006. Neural internet: warning. IEEE Trans. Biomed. Eng. 55 (5), 1582–1591.
web surfing with brain potentials for the completely paralyzed. Lin, C.-T., Wang, Y.-K., Fang, C.-N., Yu, Y.-H., King, J.-T., 2015. Extracting
Neurorehabil. Neural Repair 20 (4), 508–515. patterns of single-trial EEG using an adaptive learning algorithm. In:
Kaufmann, T., Schulz, S.M., Grunzinger, C., Kubler, A., 2012. Flashing 2015 37th Annual International Conference of the IEEE Engineering
characters with famous faces improves ERP-based brain-computer in Medicine and Biology Society (EMBC). IEEE, pp. 6642–6645.
interface performance. J. Neural Eng. 8 (5), 056016. Loew, A., Bradley, M.M., Lang, P.J., 2013. Perceptual processing of
Kaufmann, T., Schulz, S.M., Kueblitz, A., Renner, G., Wessig, C., natural scenes at rapid rates: effects of complexity, content, and
Kuebler, A., 2013. Face stimuli effectively prevent brain-computer emotional arousal. Cogn. Affect. Behav. Neurosci. 13 (4), 860–868.
interface inefficiency in patients with neurodegenerative disease. Available from: http://dx.doi.org/10.3758/s13415-013-0179-1.
Clin. Neurophysiol. 124 (5), 893–900. Long, C., Zhongpeng, W., Feng, H., Jiajia, Y., Hongzhi, Q., Peng, Z.,
Kim, Y.-H., You, S.H., Ko, M.-H., Park, J.-W., Lee, K.H., Jang, S.H., Yoo, Baikun, W., Dong, M., 2015. An online hybrid brain-computer inter-
W.-K., Hallett, M., 2006. Repetitive transcranial magnetic stimula- face combining multiple physiological signals for webpage browse.
tion-induced corticomotor excitability and associated motor skill Conf. Proc. IEEE Eng. Med. Biol. Soc. 2015, 1152–1155.
acquisition in chronic stroke. Stroke 37 (6), 1471–1476. Lopez-Larraz, E., Trincado-Alonso, F., Rajasekaran, V., Perez-
Kirchner, E.A., Kim, S.K., Tabie, M., Wohrle, H., Maurus, M., Kirchner, Nombela, S., Del-Ama, A.J., Aranda, J., Minguez, J., Gil-Agudo,
F., June 2016. An intelligent man-machine interface-multi-robot A., Montesano, L., August 3, 2016. Control of an ambulatory exo-
control adapted for task engagement based on single-trial detect- skeleton with a brain-machine interface for spinal cord injury gait
ability of P300. Front. Hum. Neurosci. 21 (10), 291. http://dx.doi. rehabilitation. Front. Neurosci. 10, 359. http://dx.doi.org/10.3389/
org/10.3389/fnhum.2016.00291. eCollection 2016. fnins.2016.00359. eCollection 2016.
Kober, S.E., Schweiger, D., Witte, M., Reichert, J.L., Grieshofer, P., Neuper, Luu, T.P., He, Y., Brown, S., Nakagame, S., Contreras-Vidal, J.L., 2016.
C., Wood, G., December 2015. Specific effects of EEG-based neurofeed- Gait adaptation to visual kinematic perturbations using a real-time
back training on memory functions in post-stroke victims. J. Neuroeng. closed-loop brain-computer interface to a virtual reality avatar. J.
Rehabil. 1 (12), 107. http://dx.doi.org/10.1186/s12984-015-0105-6. Neural Eng. 13 (3), 036006.
Pfurtscheller, G., 1981. Central beta rhythm during sensorimotor activi- Schalk, G., Wolpaw, J.R., McFarland, D.J., Pfurtscheller, G., 2000.
ties in man. Electroencephalogr. Clin. Neurophysiol. 51 (3), 253–264. EEG-based communication: presence of an error potential. Clin.
Pineda, J.A., Carrasco, K., Datko, M., Pillen, S., Schalles, M., April 28, Neurophysiol. 111 (12), 2138–2144.
2014. Neurofeedback training produces normalization in behav- Schalk, G., McFarland, D.J., Hinterberger, T., Birbaumer, N., Wolpaw,
ioural and electrophysiological measures of high-functioning J.R., 2004. BCI2000: a general-purpose brain-computer interface
autism. Philos. Trans. R. Soc. Lond. B Biol. Sci. 369 (1644), 20130183. (BCI) system. IEEE Trans. Biomed. Eng. 51 (6), 1034–1043.
http://dx.doi.org/10.1098/rstb.2013.0183. Print 2014. Scherer, R., Proell, M., Allison, B., Mueller-Putz, G.R., 2012. New input
Pohlmeyer, E.A., Wang, J., Jangraw, D.C., Lou, B., Chang, S.F., Sajda, modalities for modern game design and virtual embodiment. In:
P., 2011. Closing the loop in cortically-coupled computer vision: a 2012 IEEE Virtual Reality Workshops (VRW). IEEE, pp. 163–164.
brain-computer interface for searching image databases. J. Neural Schmorrow, D., Stanney, K.M., Wilson, G., Young, P., 2006. Augmented
Eng. 8 (3), 036025. cognition in human-system interaction. In: Handbook of Human
Polich, J., 2004. Clinical application of the P300 event-related brain Factors and Ergonomics, third ed., pp. 1364–1383.
potential. Phys. Med. Rehabil. Clin. N. Am. 15 (1), 133–161. Schnuerer, A., Espinosa, A., Guger, C., 2015. mindBEAGLE: a BCI for
Pons, J., Torricelli, D., 2014. Emerging Therapies in Neurorehabilitation. communication and assessment of consciousness for patients with
Springer. disorders of consciousness. Clin. Neurophysiol. 126 (8), e152–e153.
Prinzel, L.J., Freeman, F.G., Scerbo, M.W., Mikulka, P.J., Pope, A.T., Sellers, E.W., Kubler, A., Donchin, E., 2006. Brain-computer interface
2000. A closed-loop system for examining psychophysiological research at the University of South Florida cognitive psychophysi-
measures for adaptive task allocation. Int. J. Aviat. Psychol. 10 (4), ology laboratory: the P300 speller. IEEE Trans. Neural Syst. Rehabil.
393–410. Eng. 14 (2), 221–224.
Prinzel III, L.J., Pope, A.T., Freeman, F.G., 2002. Physiological self- Severens, M., Van der Waal, M., Farquhar, J., Desain, P., 2014.
regulation and adaptive automation. Int. J. Aviat. Psychol. 12 (2), Comparing tactile and visual gaze-independent brain-computer
179–196. interfaces in patients with amyotrophic lateral sclerosis and healthy
Ramos-Murguialday, A., Schurholz, M., Caggiano, V., Wildgruber, users. Clin. Neurophysiol. 125 (11), 2297–2304.
M., Caria, A., Hammer, E.M., Halder, S., Birbaumer, N., 2012. Shibasaki, H., Hallett, M., 2006. What is the Bereitschaftspotential?
Proprioceptive feedback and brain computer interface (BCI) Clin. Neurophysiol. 117 (11), 2341–2356.
based neuroprostheses. PLoS One 7 (10), e47048. http://dx.doi. Siebner, H.R., Tormos, J.M., Ceballos-Baumann, A.O., Auer, C., Catala,
org/10.1371/journal.pone.0047048. Epub October 5, 2012. M.D., Conrad, B., Pascual-Leone, A., 1999. Low-frequency repeti-
Reissland, J., Zander, T.O., 2009. Automated detection of bluffing in a tive transcranial magnetic stimulation of the motor cortex in writ-
game — revealing a complex covert user state with a passive BCI. er’s cramp. Neurology 52 (3), 529–537.
In: Proceedings of the Human Factors and Ergonomics Society Silverman, D., Masland, W.S., Rodin, E.A., 1968. Electroencephalography.
Europe Chapter Annual Meeting, Linkoeping, Sweden. Prog. Neurol. Psychiatry 23, 303–337.
Remsik, A., Young, B., Vermilyea, R., Kiekhoefer, L., Abrams, J., Sitaram, R., Caria, A., Veit, R., Gaber, T., Ruiz, S., Birbaumer, N., 2014.
Evander Elmore, S., Schultz, P., Nair, V., Edwards, D., Williams, Volitional control of the anterior insula in criminal psychopaths
J., Prabhakaran, V., 2016. A review of the progression and future using real-time fMRI neurofeedback: a pilot study. Front. Behav.
implications of brain-computer interface therapies for restoration Neurosci. 8, 344.
of distal upper extremity motor function after stroke. Expert Rev. Soekadar, S.R., Birbaumer, N., Slutzky, M.W., Cohen, L.G., 2015. Brain-
Med. Devices 13 (5), 445–454. machine interfaces in neurorehabilitation of stroke. Neurobiol. Dis.
Riccio, A., Pichiorri, F., Schettini, F., Toppi, J., Risetti, M., Formisano, 83, 172–179.
R., Molinari, M., Astolfi, L., Cincotti, F., Mattia, D., 2016. Interfacing St. John, M., Kobus, D.A., Morrison, J.G., Schmorrow, D., 2004.
brain with computer to improve communication and rehabilitation Overview of the DARPA augmented cognition technical integra-
after brain damage. Prog. Brain Res. 228, 357–387. tion experiment. Int. J. Hum. Comput. Interact. 17 (2), 131–149.
Rockstroh, B., Elbert, T., Birbaumer, N., Wolf, P., Duchting-Roth, A., Sterman, M.B., Mann, C.A., 1995. Concepts and applications of EEG
Reker, M., Daum, I., Lutzenberger, W., Dich-gans, J., 1993. Cortical analysis in aviation performance evaluation. Biol. Psychol. 40 (1–2),
self-regulation in patients with epilepsies. Epilepsy Res. 14 (1), 63–72. 115–130.
Rohani, D.A., Henning, W.S., Thomsen, C.E., Kjaer, T.W., Puthusserypady, Straudi, S., Fregni, F., Martinuzzi, C., Pavarelli, C., Salvioli, S., Basaglia,
S., Sorensen, H.B., 2012. BCI using imaginary movements: the simu- N., 2016. tDCS and robotics on upper limb stroke rehabilitation:
lator. Comput. Methods Prog. Biomed. 111 (2), 300–307. effect modification by stroke duration and type of stroke. Biomed.
Rohm, M., Schneiders, M., Muller, C., Kreilinger, A., Kaiser, V., Muller- Res. Int. 2016.
Putz, G.R., Rupp, R., October 2013. Hybrid brain-computer inter- Tong, J., Lin, Q., Xiao, R., Ding, L., 2016. Combining multiple features
faces and hybrid neuroprostheses for restoration of upper limb for error detection and its application in brain-computer interface.
functions in individuals with highlevel spinal cord injury. Artif. Biomed. Eng. Online 15, 17.
Intell. Med. 59 (2), 133–142. http://dx.doi.org/10.1016/j.art- Townsend, G., LaPallo, B.K., Boulay, C.B., Krusienski, D.J., Frye, G.E.,
med.2013.07.004. Epub September 13, 2013. Hauser, C.K., Schwartz, N.E., Vaughan, T.M., Wolpaw, J.R., Sellers,
Rohm, M., Schneiders, M., Muller, C., Kreilinger, A., Kaiser, V., Muller- E.W., 2010. A novel P300-based brain-computer interface stimulus
Putz, G.R., Rupp, R., 2014. Hybrid brain- computer interfaces and presentation paradigm: moving beyond rows and columns. Clin.
hybrid neuroprostheses for restoration of upper limb functions in Neurophysiol. 121 (7), 1109–1120.
individuals with highlevel spinal cord injury. Artif. Intell. Med. 59 Ushiba, J., Soekadar, S., 2016. Brain-machine interfaces for rehabilita-
(2), 133–142. tion of poststroke hemiplegia. Prog. Brain Res. 228, 163–183.
Rossi, S., Rossini, P.M., 2004. TMS in cognitive plasticity and the poten- Vaughan, T.M., McFarland, D.J., Schalk, G., Sarnacki, W.A., Krusienski,
tial for rehabilitation. Trends Cogn. Sci. 8 (6), 273–279. D.J., Sellers, E.W., Wolpaw, J.R., 2006. The Wadsworth BCI research
Rupp, R., 2014. Challenges in clinical applications of brain computer inter- and development program: at home with BCI. IEEE Trans. Neural
faces in individuals with spinal cord injury. Front. Neuroeng. 7, 38. Syst. Rehabil. Eng. 14 (2), 229–233.
Sabathiel, N., Irimia, D.C., Allison, B.Z., Guger, C., Edlinger, G., Vecchiato, G., Astolfi, L., De Vico Fallani, F., Toppi, J., Aloise, F., Bez, F.,
2016. Paired associative stimulation with brain- computer inter- Wei, D., Kong, W., Dai, J., Cincotti, F., Mattia, D., Babiloni, F., 2011.
faces: a new paradigm for stroke rehabilitation. In: International On the use of EEG or MEG brain imaging tools in neuromarketing
Conference on Augmented Cognition. Springer, pp. 261–272. research. Comput. Intell. Neurosci. 2011, 643489.