DOI 10.

1007/s11055-018-0668-3
Neuroscience and Behavioral Physiology, Vol. 48, No. 9, November, 2018

Use of Robotic Devices in Post-Stroke Rehabilitation

A. A. Frolov,1,2 I. B. Kozlovskaya,3
E. V. Biryukova,1,2 and P. D. Bobrov1,2 UDC 612.821

Translated from Zhurnal Vysshei Nervnoi Deyatel’nosti imeni I. P. Pavlova, Vol. 67, No. 4, pp. 394–413,
July–August, 2017. Original article submitted March 3, 2017. Accepted May 3, 2017.

This review addresses the use of robotic devices in the rehabilitation of poststroke and posttrauma patients
as a rehabilitation technology which has developed rapidly over the last decade. The types of devices used
are described – manipulators and exoskeletons – along with the clinical protocols for their use and the ef-
fectiveness of rehabilitation procedures. Particular attention is paid to the neurophysiological basis of the
rehabilitation potential of this technology, including analysis of measures of plastic rearrangements of the
brain. Results obtained from state-of-the-art rehabilitation technology using hand exoskeletons controlled
by brain–computer interfaces based on kinesthetic motor imagery are considered.
Keywords: neurorehabilitation, stroke, robotic devices, brain–computer interface.

Introduction. Robotic devices driving human limb
movements in response to biological signals are increas-
ingly being introduced into clinical practice. Two types of
application for these devices can be discriminated – social
rehabilitation and neurorehabilitation. Social rehabilitation
aims to involve the person in everyday life, including self-
care at home, and in interactions with people and various
appliances. Social rehabilitation makes use of prostheses
and manipulators supporting the main daily needs. The aims
of these devices are to substitute for rather than to restore
losts motor functions.
In contrast to social rehabilitation, neurorehabilitation
aims to restore motor functions lost after stroke or brain in-
jury [Chernikova, 2016]. Tools and approaches to motor re-
habilitation include therapeutic physical exercise, massage,
and physiotherapy – which have long remained limited and
insufficiently effective. The situation changed dramatical-
ly at the turn of the century as a result of the formulation
of the principles of neurorehabilitation, which are based
1 Institute of Higher Nervous Activity and Neurophysiology,
Russian Academy of Sciences, Moscow, Russia;
e-mail: ebiryukova@mail.ru.
2 Pirogov Russian National Research Medical University,
Moscow, Russia.
3 Institute of Medical-Biological Problems, Russian Academy
of Sciences, Moscow, Russia.
on the neuroplasticity of the brain [Nudo et al., 1996; Taub
et al., 2002; Bach-Y-Rita, 2001]. The studies reported by
Nudo et al. [1996] using monkeys with experimental stroke
demonstrated the efficacy of restoring MF depending on
the intensity and repetition of the goal-directed movements.
Taub et al. [2002] presented results obtained by restriction
therapy – stimulation of the targeted repeated movements of
the afflicted limb by restricting the movements of the intact
limb. Bach-Y-Rita [2001] summarized the physiological
grounds for motivated, functionally significant active train-
ing allowing progression from compensation of MF to its
restoration. Long before the principles neurorehabilitation
based on brain plasticity were formulated, functionally sig-
nificant movements were used for motor rehabilitation on
another theoretical basis – switching of the nervous system
levels at which movements are constructed using changes
in the sense of the motor task [Leont’ev and Zaporozhets,
1945; Bernshtein, 1947]. The main hypothesis underlying
both approaches is that functional plasticity comes from the
structural plasticity of the brain.
The following neurorehabilitation principles were for-
mulated on the basis of these theoretical grounds:
– active involvement of the patient in rehabilitation
procedures;
– goal-directed training to movements which are func-
tionally significant for the patient;
– intensity and regulatory of sessions.

1053
0097-0549/18/4809-1053 ©2018 Springer Science+Business Media, LLC
1054
Fig. 1. Exoskeletons produced in Russia: A) Z-hand driven by pneumatic muscles. Made by Neurobotics, Zelenograd. B) Pianist hand exoskeleton driven by
electric motors. Made by NPO Android Technics, Magnitogorsk. C) Arm exoskeleton with seven degrees of freedom (flexion-extension and abduction-ad-
duction of wrist, flexion-extension and pronation-supination of forearm, flexion-extension, abduction-adduction, and rotation relative to long axis at the
shoulder joint) driven by electric motors. Made by NPO Android Technics, Magnitogorsk.
Rehabilitation programs using robotic devices pro-
vides ideal realization of these principles. Firstly, these
devices produce many repetitions of movements at a con-
trolled speed, “without producing fatigue,” in contrast to
work with medical staff. Secondly, the repertoire of move-
ments most appropriate to the individual characteristics of
the patient can be selected [Wolbrecht et al., 2006]. Robot
devices are fitted with sensors detecting the kinetic and
force parameters of movements, providing feedback and
objective evaluation of movement parameters [Reiner et
al., 2005]. Feedback based on detection of movement pa-
rameters allows a virtual environment to be generated on a
computer which is either near-real or includes gaming com-
ponents, which increases motivation to take part in rehabil-
itation [Jack et al., 2001].
As a rule, the use of robotic systems requires some
preserved movements. If a movement is completely absent
(plegia) or is too minimal, the only means of activating the
plastic mechanisms of the brain for rehabilitation consists
of kinesthetic motor imagery. Motor imagery is known to
activate essentially the same areas of the brain as execution
of movements [Jeannerod and Frak, 1999]. Signals relating
to the activity of these areas during motor imagery can be
transformed into commands controlling robotic systems via
a BCI. Systematic controlled studies of the effectiveness of
rehabilitation procedures using robotic systems controlled
by BCI started to be undertaken about ten years ago, but are
still few in number [Buch et al., 2008; Ang et al., 2014, 2015;
Ramos-Murguialday et al., 2013; Ono et al., 2014; Kotov et
al., 2014; Mokienko et al., 2016; Frolov et al., 2016a].
In the present article:
1. Robotic systems used in neurorehabilitation are
reviewed;
2. Methods for analysis of plastic rearrangements of
the brain are described and some results obtained using
Frolov, Kozlovskaya, Biryukova, and Bobrov
these methods in both healthy subjects and patients under-
going training to control a BCI are presented;
3. Clinical data demonstrating the effectiveness of pro-
cedures using hand exoskeletons controlled by BCI for res-
toration of MF in poststroke patients are presented.
Robotic Devices Used in Neurorehabilitation
1. Types of robotic devices. Robotic devices currently
used in neurorehabilitation can be divided into two groups
– manipulators and exoskeletons. Manipulators are attached
only to the distal part of the arm – the hand or fingers. As a
rule, the patient holds the handle of the manipulator, which
carries out some movement or other. Thus, a manipulator
controls the operating point of the hand but not its individual
joints. In cases of severe movement disorders, movements at
the arm joints follow the hand passively, decreasing rehabil-
itation effectiveness [Marchal-Crespo and Reinkensmeyer,
2009]. In cases of moderate motor disorders, when there is
persistence, perhaps to a limited extent, of a) the ability of
the hand to hold the handle and b) movement at the elbow
and shoulder joints, manipulators can operate in assistance
mode using the residual muscle activity.
Exoskeletons directly controlling movements at the
joints of the arm constitute a more physiologically appro-
priate means of rehabilitation than manipulators, at least for
severe motor disorders. However, active exoskeletons are
much more complex technical devices than manipulators:
the exoskeleton structure must adapt to the anatomical pe-
culiarities of the patient’s arm – the lengths of the segments
and the positions of the axes of rotation of the joints; control
of torque must be concordant with the viscoelastic proper-
ties of the neuromuscular apparatus.
Manipulators for neurorehabilitation started to be de-
veloped at the end of the 1990s, some 10 years earlier than
exoskeletons. There are now commercial versions active-
ly used in poststroke rehabilitation – InMotion ARM and
Use of Robotic Devices in Post-Stroke Rehabilitation 1055

TABLE 1. Manipulators for Motor Recovery of Arm Movements Tested in Controlled Clinical Trials (F-E – flexion-extension, Ab-Ad – abduction-adduction,
P-S – pronation-supination, Rot – rotation about the longitudinal axis)

Description, clinical
Developer Country Manipulator actions Arm joints
application
Interactive Motions [Ang et al., 2011,
MIT-Manus Damps deviations from speci- Shoulder, elbow (move-
Technologies USA, USA 2015; Krebs et al.,
(InMotion ARM) fied trajectory ment in specified plane)
Watertown, MA 1998]
T-WREX (Therapy National Institute of
Assistance mode in computer Shoulder (F-E, Ab-Ad,
Wilmington Robotic Disability and USA [Sanchez et al., 2004]
games Rot), elbow (F-E)
Exoskeleton) Rehabilitation Research
MIME (Mirror Shoulder, elbow (rectilin-
The Catholic University Reproduces trajectory of
Image Movement USA [Lum et al., 2006] ear movements of forearm
of America healthy arm for paretic arm
Enabler) in different directions)
Reha-Stim Medtec GmbH Active and passive bimanual Elbow (P-S), radiocarpal
Bi-Manu-Track Germany [Hesse et al., 2003]
& Co. KG modes (F-E)
NeReBot
Recording and reproduction of
(NEuro- Shoulder (Ab-Ad), elbow
Padua University Italy [Rosati et al., 2007] arm trajectory recommended
REhabilitation- (F-E, P-S)
by rehabilitation physician
RoBOT)
Execution of a large repertoire
of slow movements at constant Shoulder (Ab-Ad, F-E,
REHAROB European project – REHAROB [Fazekas et al., 2006] speed to increase movement Rot), elbow (Ab-Ad, F-E),
amplitude and decrease joint radiocarpal (Ab-Ad, F-E)
spasticity
Shoulder (Ab-Ad, F-E),
[Amirabdollahian Operates in three modes: pas-
Haptic Master European project – Gentle/G elbow (F-E, P-S), radiocar-
et al., 2007] sive, assistance, and active
pal (F-E)
MEMOS
Assistance mode in the process
(MEchatronic system Scuola Superiore Shoulder (Ab-Ad, F-E),
Italy [Micera et al., 2005] of moving the patient’s arm to
for MOtor recovery Saint’Anna in Pisa elbow (F-E)
a specified target
after Stroke)
Assistance mode in the process
Rehabilitation Institute of [Reikensmeyer et al., Shoulder (F-E), elbow
ARM Guide USA of moving the patient’s arm to
Chicago 2000] (F-E)
a specified target

ArmeoSpring, based, respectively, on the original devel-
opments MIT-Manus and T-WREX (Table 1). Along with
MIT-Manus and T-WREX, Table 1 shows other manipula-
tors which have been used in controlled clinical trials.
The development of exoskeletons over the last 10 years
has run at such a high rate that the number is currently com-
parable with the number of manipulators developed. For ex-
ample, a detailed review [Maciejasz et al., 2014] listed 44
manipulators and 35 exoskeletons intended for poststroke
rehabilitation. Most exoskeletons have been developed for
the hand. There are at least two reasons for this. First, the
hand and fingers have the greatest level of impairment after
stroke and recovery of their movements is prolonged and dif-
ficult [Shelton and Reding, 23001]. Second, the arrangement
of the muscular apparatus of the fingers can be reproduced
quite simply for exoskeletons: the fingers extend by means of
the common extensor whose individual tendons are attached
to the distal phalanges, and they flex by means of common
flexors. The muscles controlling flexion and extension of the
thumb also attach to the distal phalanges. The contractile
parts of all these muscles are located in the forearm, only the
slender tendons running into the hand [Sinel’nikov, 1967].
Thus, an artificial pair of antagonistic muscles located out-
side the fingers can provide anatomically appropriate control
of flexion-extension of the fingers using a mobile exoskele-
ton attached to the distal phalanges (Fig. 1).
Most exoskeletons are at the prototype stage and have
been approved only in healthy subjects [Bos et al., 2016; Heo
et al., 2012]. However, exoskeleton constructions for post-
stroke patients have much more stringent requirements than
those for healthy subjects. These restrictions are mainly as-
sociated with arm and hand spasticity, which hinders attach-
ment to the exoskeleton. Exoskeletons tested on poststroke
patients are shown in Table 2 (hand) and Table 3 (arm).
A total of three hand exoskeletons – HWARD, Reha-
Digit, and Hand Mentor (Table 2) – and two arm exoskele-
tons – T-WREX and ARMOR (Table 3) – have been used in
controlled clinical trials in which the effectiveness of proce-
dures using exoskeletons were compared with the effective-
ness of standard poststroke treatment.
Among hand exoskeletons approved in clinical tri-
als, those in Table 2 include two Russian-made models:
1056 Frolov, Kozlovskaya, Biryukova, and Bobrov

TABLE 2. Hand and Finger Exoskeletons

Developer Country Description Action of exoskeleton Arm joints

HWARD (Hand Flexion-extension of hand
Radiocarpal, 1st carpo-
Wrist Assistive at radiocarpal joint, thumb
University of California USA [Takahashi et al., 2008] metacarpal, 2nd–5th
Rehabilitation and fingers 2–5
metacarpophalangeal
Device) simultaneously
[Kotov et al., 2014;
Hand exoskeleton Simultaneous extension of
Neurobotics, Zelenograd Russia Mokienko et al., 2016; All joints of fingers 1–5
Z-hand fingers
Frolov et al., 2016]
Hand exoskeleton NPO Android Technics, Simultaneous extension of Radiocarpal, all joints of
Russia [Kondur et al., 2016]
Pianist Magnitogorsk fingers fingers 2–5
Hand of Hope
Rehab-Robotics Comp. Hong Kong, Independent flexion-exten-
(commercial [Ho et al., 2011] All joints of fingers 2–5
Ltd. China sion of fingers 2–5
system)
Active linear
National University of
HandCARE Singapore [Dovat et al., 2008] displacement of fingers All joints of fingers 2–5
Singapore
2–5
Center for Applied
HEXORR (Hand
Biomechanics and Simultaneous flexion-ex-
EXOskeleton [Schabowsky et al.,
Rehabilitation Research, USA tension of fingers 2–5, All joints of fingers 1–5
Rehabilitation 2010]
National Rehabilitation flexion-extension of thumb
Robot)
Hospital, Washington
InMotion Hand
Interactive Motion Tech.,
(commercial USA [Masia et al., 2007] Ball grip All joints of fingers 1–5
Inc.
system)
Electronic and Information Simultaneous flexion-ex-
Hand exoskeleton Department, Politecnico Italy [Mulas et al., 2005] tension of fingers 2–5, All joints of fingers 1–5
di Milano flexion-extension of thumb
Simultaneous flexion-ex-
Reha-Digit,
tension of fingers 2–5,
Reha-Stim Reha-Stim Medtec, GmbH
Germany [Hesse et al., 2008] vibrational stimulation of All joints of fingers 2–5
(commercial & Co, Berlin
the palmar surface of the
system)
fingers used in plegias
Department of Health
Technology and Metacarpophalangeal and
Independent flexion-exten-
Hand exoskeleton Informatics, The Hong China [Tong et al., 2010] distal interphalangeal
sion of fingers 2–5
Kong Polytechnic joints of fingers 2–5
University
AMES (Assisted Simultaneous flexion-ex-
AMES Technology, Inc.,
Movement with tension of hand and fingers Radiocarpal and metacar-
a spin-off company from
Enhanced USA [Cordo et al., 2009] 2–5, at the metacarpopha- pophalangeal joints of
Oregon Health & Science
Sensation) (com- langeal joints, vibrational fingers 2–5
University
mercial system) stimulation of muscles
Hand Mentor™
Kinetic Muscles, Inc., Extension of hand and Radiocarpal and joints of
(commercial USA [Koeneman et al., 2004]
Tempe, Arizona fingers 2–5 fingers 2–5
system)
AMADEO robotic [Kwakkel, Meskers, Flexion-extension of each
Tyromotion GmbH Graz Austria All joints of fingers 1–5
system 2014] finger separately

the Z-hand exoskeleton, operated by pneumatic muscles
(Neurobotics, Fig. 1, A), and the Pianist exoskeleton, oper-
ated by electric motors (NPO Android Technics; Fig. 1, B).
These exoskeletons have been used in a controlled, multi-
center, randomized trial [Frolov et al., 2016a].
2. Biologically appropriate means of controlling ro-
botic devices. A fundamental challenge for the use of ma-
nipulators and exoskeletons in rehabilitation is that of con-
trolling these devices, which should ideally follow the prin-
ciples of controlling the human hand by the CNS. Otherwise,
the patient perceives a movement made by an external de-
vice as imposed and unfriendly, which leads to a decrease in
the effectiveness of rehabilitation. Improvements in means
of controlling robotic devices are needed because of insuffi-
cient understanding of the principles of movement control
by the CNS [Ferris, 2009]. However, the achievement of
Use of Robotic Devices in Post-Stroke Rehabilitation 1057

TABLE 3. Exoskeletons Supporting Movements at Different Joints of the Arm (F-E – flexion-extension; Ab-Ad – abduction-adduction, P-S – pronation-su-
pination, Rot – rotation relative to the longitudinal axis of the arm)

Developer Country Description Action of exoskeleton Arm joints

MUNDUS
Shoulder (Ab-Ad, F-E),
(MUltimodal Italy, Austria,
European project [Pedrocchi et al., Assistance mode in every- elbow (F-E, P-S), radiocar-
Neuroprosthesis for Germany,
MUNDUS 2013] day actions pal (F-E), finger joints
Daily Uppe limb Switzerland
(ball grip)
Support)
Shoulder (Ab-Ad, F-E),
Assistance mode stimulat-
Sensory Motor System elbow (F-E, P-S), radiocar-
ARMin Switzerland [Nef et al., 2009] ing involvement of the
Lab, Universität Zürich pal (F-E), finger joints
patient
(ball grip)
RUPERT (Robotic
Shoulder (F-E, Rot), elbow
UPper Extremity [Balasubramanian Assistance mode in every-
Arizona State University USA (F-E, P-S), radiocarpal
Repetitive Therapy et al., 2008] day actions
(F-E)
device)
T-WREX (Training Department of Mechanical Shoulder (Ab-Ad, F-E,
Wilmingyon Robotic & Aerospace Engineering, USA [Sanchez ey al., 2004] Passive mode Rot), elbow (F-E), finger
EXoskeleton) California University joints (ball grip)
Shoulder (Ab-Ad, F-E,
PneuWREX Minimum required assis-
[Wolbrecht et al., horizontal movement),
(Wilmington Robotic USA tance mode in everyday
2006] elbow (F-E), finger joints
EXoskeleton) actions
(ball grip)
Provides feedback from
PERCRO Laboratory,
L-exos (Light pressure on the palmar Shoulder (Ab-Ad, F-E,
Scuola Superiore Italy [Frisoli et al., 2007]
exoskeleton) surface of the hand during Rot), elbow (F-E)
Sant’Anna, Pisa
training movements
Design and Materials
ARAMIS Two symmetrical exoskel-
Technology Centre,
(Automatic etons, one of which com-
Brindisi; S. Anna Institute Shoulder (Ab-Ad, F-E,
Recovery Arm Italy [Pignolo et al., 2012] pensates for lack of
and Research on Advanced Rot), elbow (F-E, P-S)
Motility Integrated strength and accuracy in
Neurorehabilitation,
System) the paretic arm
Crotone
Department of
Passive mode, restricting
Biomechanical Shoulder (Ab-Ad, F-E,
Dampace Netherlands [Stienen et al., 2009] movement in specific de-
Engineering, University Rot), elbow (F-E)
grees of freedom
of Twente, Enschede
Shoulder (Ab-Ad, F-E,
Abteilung für
Rot), elbow (F-E, P-S),
ARMOR Neurologische Austria [Mayr et al., 2008] Active training mode
thumb (F-E), fingers 2-5
Akutbehandlung
(simultaneous F-E)
Shoulder (Ab-Ad, F-E,
Rehabilitation Institute Active mode: assistance or Rot), elbow (F-E), radio-
IntelliArm USA [Ren et al., 2009]
of Chicago preventative carpal (F-E), finger joints
(ball grip)

physiological movements nonetheless provides a theoreti-
cal basis for developing biologically appropriate means of
control.
Theoretical and experimental studies have shown that
planning of goal-directed arm movements occurs in terms
of the trajectory of the working point in external space, for
example, the trajectory of the hand for grasping an object
or the trajectory of the tip of the finger for placing it at a
specific point [Hogan and Flash, 1987]. Thus, manipulators
forming the trajectory of the working point of the arm carry
out the plan of a goal-directed movement in a physiologi-
cally appropriate manner. However, studies have shown that
damage to brain structures generally produces less impair-
ment to planning of movements in terms of the trajectory
of the working point than to the mechanisms of interactions
between joints [Levin, 1996]. This raises the challenge of
the physiologically appropriate formation of joint rotations.
The task of transforming the trajectory of the working point
to joint rotations does not have any single solution because
of the overprovision by the skeletal and neuromuscular ap-
paratus of the human arm. This transformation is learned
during the process of motor training, leading to the forma-
1058
tion of synergies of central commands, which are apparent
as coordinated changes in joint angles (kinematic syner-
gies) and muscular torque at the joints (dynamic synergies)
[Bernshtein, 1935].
There are currently no generally accepted methods for
determining motor synergies suitable for use in active exo-
skeletons. Hand exoskeletons (Table 2) do not need joint
rotations for the trajectory of the working point. The con-
struction of these exoskeletons is close to the anatomical
construction of the flexor-extensor apparatus of the fingers
and provide good reproduction of the main grasping move-
ments of the fingers. Arm exoskeletons (Table 3) are also
not faced with the task of realizing any particular kinematic
or dynamic synergies. As a rule, an assistance mode is used,
where residual movements in the arm joints are amplified
by feedback signals, or in the passive mode, where move-
ments are restricted into one or another set of degrees of
freedom, in accordance with the rehabilitation protocol. Just
one arm exoskeleton approved in clinical trials – ARMOR
[Mayr et al., 2008] – reproduces, in the active mode, the
joint movements of the functionally intact arm in the joints
of the paralyzed arm.
Some authors have suggested that in transforming the
trajectory of the working point into changes in joint angles,
the nervous system solves several optimization tasks. This
approach, actively developed in the 1970s–1980s [Bizzi et
al., 1982], also has more contemporary supporters [Kaneko
et al., 2005]. This suggestion, which is entirely logical in
prototyping robotic system, is rather artificial for living sys-
tems. A more biologically appropriate approach consists of
using a controlling neural network which learns on the basis
of trial and error [Frolov et al., 1994]. Both approaches, al-
though supporting selection of a trajectory in the space of
the joint angles close to the experimentally observed, are
not universal because of the arbitrariness of selecting the
target function to be optimized and because of the neural
network training algorithm.
Another complexity in controlling exoskeletons comes
from the viscoelastic nature of the joint torque in combina-
tion with low physiological joint stiffness levels. In living
systems, this difficulty is overcome by control using feed-
back and multiple repetition of movements during motor
learning [Bernshtein, 1947/1991]. Given that joint stiffness
in the arm is not great (according to different literature sourc-
es: from 2 to 100 Nm at the elbow and from 10 to 70 Nm at
the shoulder) [Frolov et al., 2000, 2006], rigid control of the
joint in executing a specified kinematic is perceived by the
patient as uncomfortable. Biologically appropriate control
of the joint torque requires feedback from the joint angle and
for the exoskeleton to provide the low joint stiffnesses that
the patient is familiar with. Feedback from the joint angle
and control in terms of impedance, equivalent to control us-
ing a PD (proportional derivative) – the controller in techni-
cal systems – is successfully used in manipulators (Table 1)
and exoskeletons (Tables 2, 3). The review of Anam and
Frolov, Kozlovskaya, Biryukova, and Bobrov
Al-Jumailyb [2002] provides a detailed description of the
feedback control of the REHAROB manipulator and the
IntelliArm, ARMin, L-Exos, Rupert, and Pneu-WREX arm
exoskeletons.
However, the problem of low stiffnesses in exoskel-
eton articulations remains unsolved. Experimental physi-
ological studies have demonstrated that control of a joint
with low stiffness comes up against the problem of control
stability. In living systems, the total torque at a joint can be
modeled by a viscoelastic spring with controllable parame-
ters: equilibrium length, stiffness, and viscosity [Fel’dman,
1979; Gomi and Kawato, 1996]. The equilibrium length of a
spring corresponds to the desired joint angle and is specified
at the supraspinal level of the CNS. Stiffness and viscosity
are determined by the interaction of three different physio-
logical mechanisms: the viscoelastic properties of muscle
fibers and tendons, the spinal stretch reflex circuit, and the
feedback circuit passing through the brain [Peterka, 2002].
In contrast to the first mechanism, operation of the second
and third have significant delays. The delay for the stretch
reflex circuit is ~50 msec, compared with up to 150 msec for
the second circuit [Peterka et al., 2002; Frolov et al., 2000,
2006]. The effective delay used in feedback control is de-
termined by the relative contributions of these three mech-
anisms. The existence of a time delay imposes restrictions
on the feedback parameters supporting stable control. These
restrictions are apparent in that stiffness and viscosity must
be limited both from above and from below [Alexandrov et
al., 2005].
The appropriateness of the model of control of the total
muscular torque at the joint based on a viscoelastic spring
with controllable parameters has been demonstrated for
control of the elbow joint in conditions of unexpected con-
trolled unloading of the forearm [Biryukova et al., 1999], by
the joints of the human arm in goal-directed movements
[Frolov et al., 2006], by the hip, knee, and ankle joints when
people are tilted in the sagittal plane [Alexandrov et al.,
2005; Alexandrov and Frolov, 2011].
An effective means of computing biologically appro-
priate control parameters using feedback based on the prin-
ciple of independent control of movements along dynamic
control eigenvectors has been proposed in [Kaneko et al.,
2005; Alexandrov and Frolov, 2011; Frolov et al., 2013].
This has been realized for control of an arm exoskeleton
with seven degrees of freedom made by NPO Android
Technics (Magnitogorsk) and has been tested in healthy
subjects (Fig. 3).
3. Neurorehabilitation technologies. Technical solu-
tions realizing active actions by the exoskeleton on the pa-
tient’s arm during neurorehabilitation can be divided into
assistance, restrictive, and technologies using BCI.
3.1. Assistance technologies. Impedance-assistance.
As long as the arm follows the specified trajectory, the ex-
ternal device does not interfere; however, if it deviates in
terms of some critical parameter, the external device devel-
Use of Robotic Devices in Post-Stroke Rehabilitation
ops a force returning the arm towards the desired trajec-
tory. In general, the returning force acts as a viscoelastic
spring [Anam and Al-Jumailyb, 2002]. To provide natural
variability in movements, some family of trajectories re-
garded as available is considered, i.e., those which when
followed do not involve the exoskeleton acting on the arm
[Reinkensmeyer et al., 2000].
Sometimes the external device interferes in the pa-
tient’s voluntary movement only when the patient cannot
complete the desired movement. In this case, the exter-
nal device helps this to happen. The MIT-Manus. MIME,
Haptic Master, and ARM Guide manipulators, the hand
exoskeleton HWARD, and the hand exoskeleton Rupert are
able to operate in this mode. The advantage of this control is
that the patient initiates the movement independently, which
is fundamental for successful motor learning [Lotze et al.,
2003]. The drawback of this control is the risk of triggering
an uncontrolled movement.
Equalizing assistance. This frequently used technolo-
gy is based on equalizing the force of gravity when the arm
rests on a support or is elevated. Some methods consider
partial unloading of the arm and use devices controlling the
degree of unloading during training [Marchal-Crespo and
Reinkensmeyer, 2009].
Assistance using EMG signals. Within the framework
of this technology, the exoskeleton assists in completing
movements, developing a force proportional to the EMG
amplitude [Ho et al., 2011; Mulas et al., 2005; Tong et al.,
2010; Koeneman et al., 2004]. This method has significant
limitations because of the sensitivity of the EMG to electrode
positioning and the overall state of the patient. In addition,
there is the risk of unpredictable exoskeleton movements
when the patient inappropriately activates muscles, which
often occurs as a result of stroke [Dipietro et al., 2005].
3.2. Restrictive technologies. Poststroke patients often
use the healthy arm or joints with retained motor functions
in executing motor tasks. The most widespread example is
movement of the body instead of extending the arm if re-
quired to reach the target. These cases use a restriction strat-
egy where the healthy arm or preserved joints are fixed with
the aim of stimulating movement in the joints with impaired
movement function [Taub et al., 2002; Shaw et al., 2005]. In
mild forms of hemiparesis, recovery mechanisms are stimu-
lated by addition of forces preventing movement and re-
quiring the patient to make an effort to overcome them
[Hesse et al., 2003; Rosati et al., 2007].
In some rehabilitation procedures, movement execu-
tion errors are increased artificially by movement perturba-
tions [Patton et al., 2006] or distortion of visual feedback
[Brewer et al., 2008]. Increases in kinematic errors (over-
shooting the target or deviation from the specified trajecto-
ry) are additional neural signals triggering adaptation mech-
anisms to correct errors.
3.3 Technologies using BCI. The effectiveness of re-
habilitation procedures depends on the extent to which they
1059
trigger the plastic mechanisms of the brain to rearrange the
sensorimotor system [Di Pino et al., 2014]. For this, exo-
skeleton movement must be carried out only at the moment
at which the brain is most sensitive to receiving peripheral
signals produced by exoskeleton movement, i.e., those mo-
ments at which the patient intends to complete the move-
ment. In many studies, the patient’s intent to carry out a
movement is recognized by detecting the EMG with super-
ficial electrodes [Mulas et al., 2005; Tong et al., 2010]. In
this case, increases in the EMG serve as a trigger signal for
the exoskeleton. However, this signal cannot be used in
completely paralyzed patients or patients with impairment
to normal muscle coactivation [Wright et al., 2014].
BCI based on kinesthetic motor imagery were used in
[Buch et al., 2008; Ang et al., 2011, 2014, 2015; Ramos-
Murguialday et al., 2013; Ono et al., 2014; Kotovet et al.,
2014; Mokienko et al., 2016; Frolov et al., 2016] to recog-
nize the patient’s intent. The patient’s intent to carry out a
movement was successfully linked with real movements
performed by the exoskeleton. Patients’ ability to lean mo-
tor imagery was comparable with that of healthy people
[Frolov et al., 2017]. It is also important that over a quite
wide range of patients, no correlations were found between
task performance success and the level of cognitive deficit
[Kotov et al., 2014; Frolov et al., 2016; Ang et al., 2011].
All the studies cited above apart from that of Ono et al.
[2014] used hand exoskeletons. The expectation that the use
of these exoskeletons will be effective is associated with the
large representation of the hand in the cerebral cortex, such
that large areas of the brain are activated. On the other hand,
movements performed by these exoskeletons – flexion/ex-
tension of the fingers – are far from the multiple-joint move-
ments performed by people in their daily lives. However, it
is well known [Bernshtein et al., 1947] that patients with
motor disorders perform highly important functional auto-
mated movements far better than purposeless movements.
In fact, the whole practice of poststroke rehabilitation is
based on the intensification and repetition of target multi-
ple-joint movements. It can be suggested that the simplicity
and absence of an aim for movements had the result that the
effects obtained using the BCI + exoskeleton procedure
were less than expected when they were first introduced
[Buch et al., 2008; Ang et al., 2011; Kotov et al., 2014]. It
therefore seems encouraging to rehabilitate using the exo-
skeleton procedure providing goal-directed movements ori-
ented to functionally significant tasks under BCI control.
The single report of studies of the potential for rehabil-
itation procedures using BCI controlling a multicomponent
arm exoskeleton providing the whole range of human arm
movements is that of Brauchle et al. [2013]. This study used
an Armeo Power, Hocoma (Volketswil, Switzerland), con-
trolling all seven degrees of freedom of the human arm. The
exoskeleton supports the arm in the gravitational field and
allows a large repertoire of goal-directed arm movements to
be made. This study demonstrated that the exoskeleton
1060
could be linked with a BCI to carry out three sequential
movements towards fixed targets in nine healthy subjects
and two poststroke patients. The effectiveness of this sys-
tem for motor rehabilitation was not assessed.
Physiological Requirements for the Use of Robotic
Devices in Poststroke Rehabilitation
1. fMRI and transcranial magnetic stimulation
data. fMRI study data from brain areas showing increases
in hemodynamic activity over the resting level on kinesthet-
ic motor imagery and on performance of movements almost
coincided [Jeannerod and Frak, 1999]. The exclusion is the
motor cortex, which is activated on execution of move-
ments but whose activity in motor imagery is seen in only
15% of studies [Hétu et al., 2013; Grèzes and Decety, 2001].
Changes in brain activity due to training to motor im-
agery in controlling a BCI were specifically studied by
Mokienko et al. [2013] and Frolov et al. [2016b] in healthy
subjects serving as the control group for comparison with
poststroke patients. These studies showed that at least 14
brain areas showed increases in levels of hemodynamic ac-
tivity due to training to motor imagery in the BCI control
procedure. However, significant increases were seen only in
the following five areas: Brodmann fields 44 and 45, the
insula, the middle frontal gyrus, and the anterior cingulate
gyrus. It is interesting that all these areas are in the anterior
part of the brain, though imagination and execution of
movements show a strong interaction between the anterior
and parietal areas of the brain [Hétu et al., 2013; Grèzes and
Decety, 2001]. The absence of any significant differences in
activity in the parietal areas of the brain appears to be asso-
ciated with the fact that the sensory flow of information in
motor imagery is limited and training to control a BCI does
not require any increase in activity in the areas processing it.
In contrast, activity in the anterior areas forming motor
commands increases.
Apart from changes in brain hemodynamic activity
on motor imagery, Mokienko et al. [2013] studied plastic
changes in the brain as a result of training to control a BCI
using transcranial magnetic stimulation. Motor imagery is
known to decrease the magnetic field intensity triggering
threshold for the activity of the relevant muscles on stimula-
tion of the primary motor cortex [Stinear, 2010]. Important
evidence for plastic rearrangements of the brain due to train-
ing to control a BCI was provided by the observation that
the increase in primary motor cortex arousability was sig-
nificantly more marked in trained than untrained subjects.
2. EEG activity data. Most reports of brain activity
during the execution, imagination, and observation of move-
ments have been obtained using fMRI studies. However, the
functional role of each of the areas whose activity increased
on execution of the tasks remained unclear. This is largely
due to the poor time resolution of fMRI – of the order of sec-
onds. EEG recording provides an alternative to fMRI and has
resolution on the order of tens of milliseconds. However, the
signal recorded by each electrode is a superimposition of the
Frolov, Kozlovskaya, Biryukova, and Bobrov
signals generated by a large set of sources, which, as demon-
strated by fMRI, are distributed through the whole volume
of the brain. This fact is responsible for the poor spatial res-
olution of the EEG. Recent years have seen the development
of an approach increasing the resolving ability of the EEG to
at least the level obtained with fMRI, retaining its high time
resolution [Makeig et al., 1996; Onton et al., 2006; Delorme
et al., 2012]. This approach is based on use of the indepen-
dent components analysis method [Kachenoura et al., 2008]
and solution of the inverse EEG problem.
This method identified tens of independent EEG sourc-
es active during motor imagery, though systematic study of
these has only just started [Frolov et al., 2012, 2013; Bobrov
et al., 2011; Bobrov et al., 2016].
Movement, preparation for movement, observation of
movement, or imagination of the movement of some execu-
tive organ is usually accompanied by decreases in the μ and
β rhythms in the cortical representation of this organ. This
decrease is termed event-related desynchronization (ERD)
[Pfurtscheller and Berghold, 1989; Pfurtscheller, 1999;
Pfurtscheller and Lopes da Silva, 1999]. Increases in the μ
rhythm, i.e., event-related synchronization (ERS) are seen
in brain areas representing organs not involved in execu-
tion of the movement of interest [Pfurtscheller, 1999]. One
hypothesis is that the ERD and ERS reactions correspond
to cortical control of the “thalamic gating mechanism”
[Klimesch et al., 2007; Klimesch, 2012]. Cortical areas
demonstrating ERS are tuned to perceive strong but unde-
fined signals without detailed encoding, while those demon-
strating ERD are tuned to the detailed analysis of specific
signals uniformly encoded over the whole region in which
they change. On execution of motor imagery too, afferent
signals (for example, proprioceptive) must constantly be
compared with efferent copies and any deviation arising as
a result of this comparison must be corrected immediately
[Jeannerod, 1994]. It can be suggested that the site at which
this comparison takes place is the primary somatosensory
cortex [Sun et al., 2015]. It can also be suggested that the
efferent copy generated by the premotor cortex arrives in
the primary somatosensory cortex and switches it to the
ERD regime, tuning to detailed comparison with peripheral
signals associated with the movement. This interpretation
of the functional significance of the ERD and ERS reac-
tions is consistent with the localization of the EEG source
demonstrating these reactions most strongly as reported in
[Frolov et al., 2012], in the depth of the central sulcus and
Brodmann area 3a, where the proprioceptive sensation of
the palms and fingers is represented.
Thus, the main, albeit not the only, neurophysiological
grounds for high operating quality of BCI based on motor
imagery are ERD and ERS reactions, i.e., suppression of the
resting μ rhythm in the representation area of the proprio-
ceptive sense of the arm whose movement is imagined, with
increases in this rhythm in the representation area of the
opposite arm.
Use of Robotic Devices in Post-Stroke Rehabilitation 1061

TABLE 4. Use of BCI + Exoskeleton Technology in Clinical Practice for Motor Recovery After Stroke

Extent of Age of Duration Assessment

Exo- Motor Number of
Description impairment to patients of stroke Protocol of Results
skeleton imagery patients
MF (years) (months) effectiveness
1. BCI +
Improvement in
Haptic Knob.
Flexion- MF in group 1
[Ang et al., Haptic F-M 10–50 2. Haptic
extension 7+8+7 30–79 6–22 F-M significantly great-
2014] Knob (max 66) Knob
of fingers er than in groups 2
3. Standard
and 3
treatment
Improvement in
1. BCI + MF in both groups
[Ang et al., MIT- Movement of F-M 4–40
11 + 15 35–62 2–18 MIT-Manus F-M with no significant
2015] Manus arm to target (max 66)
2. Mit-Manus difference between
them
Significant im-
36 (20, 13,
provement in MF
and 3 from
1. BCI + F-M, ARAT, seen in a larger
[Frolov et Extension different F-M 61–104
Z-hand 47–73 1–13.5 Z-hand MAS, proportion of pa-
al., 2016] of fingers clinics) + 11 (max 126)
2. Z-hand MRC-SS tients of the study
(control
group than the
group)
control group
1. BCI with
Improvement in
visual feed-
[Ono et al., Hand Extension SIAS0–1 SIAS, MF of fingers only
6+6 41–84 4–155 back
2014] orthosis of fingers (max 5) EMG in BCI + orthosis
2. BCI +
group
orthosis
Stretching Improvement in
[Ramos-
of arm and 1. BCI + MF in study group
Murgui- Arm F-M 4–21
grasping 16 + 16 37–62 21–111 orthosis F-M significantly great-
alday et al., orthosis (max 54)
imaginary 2. Orthosis er than in control
2013]
apple group
Flexion- 5 + 3 (two Hand plegia
[Buch et al., Wrist BCI + No improvement
extension different (MRC 0/5, 14–68 12–41 MRC
2008] orthosis orthosis in MF
of fingers clinics) MAS < 3)

Use of BCI + Exoskeleton Technology in Clinical
Studies
1. Results of controlled studies. There are as yet few
applications of exoskeleton-controlling BCI based on rec-
ognition of the EEG in kinesthetic motor imagery. Table 4
shows those which included a control group of patients not
performing a mental motor imagery task. The studies in-
cluded patients with a wide range of ages (from 14 to 84
years), durations of stroke (from one month to 13 years),
and extents of MF impairment (from 4 to 50 points out of
the maximum 66 points on the Fugl-Meyer scale). Studies
differed in terms of the numbers of procedures and their du-
rations, and procedures used different types of exoskele-
tons. Despite these differences, all studies (Table 4) showed
that improvements in MF after procedures were more
marked in the study groups than the control groups. This
leads to the conclusion that the use of BCI controlling limb
exoskeletons on the basis of EEG signals corresponding to
kinesthetic motor imagery has positive clinical effects.
It should be noted that the extent of neurological defi-
cit in these studies was mostly moderate. There is particu-
lar interest in a case of plegia and profound pareses, where
motor imagery was the only active paradigm modulating
neuroplasticity processes in the motor zones of the brain
[Mokienko et al., 2013; Shih et al., 2012; Soedakar et al.,
2015; Grosse-Wentrup et al., 2011]. Buch et al. [2008] stud-
ied eight patients with hand plegia (muscle strength on the
Medical Research Council (MRC) scale was 0/5 and muscle
spasticity on the Modified Ashworth Scale (MAS) was <3)
who underwent courses of rehabilitation with BCI-controlled
hand orthoses. Despite the fact that all patients successfully
controlled the orthosis and that the number of procedures
was significant (16–22), no improvements in MF were seen
after the procedures. It may be that this negative result was
associated with the low sensitivity of the MRC and MAS
scales used to evaluate MF. Assessment of biomechanical
parameters of the movements of patients with hand plegia,
extensive cortical injuries, and long durations of brain trau-
ma demonstrated improvements in MF after 15 BCI + hand
exoskeleton procedures [Biryukova et al., 2016].
2. Assessment of the effectiveness of BCI + exoskele-
ton technology. The question of the appropriate evaluation
1062
of MF after stroke is extremely relevant, particularly in re-
lation to assessment of new rehabilitation methods. Most
studies assessing the effectiveness of exoskeleton-con-
trolling BCI have used the Fugl-Meyer scale [Fugl-Meyer et
al., 1975], which includes a large number of motor tests and
is recommended as a validated and reliable scale [Gladstone
et al., 2002]. However. This scale, like other clinical scales,
is not free from subjective components and, using discrete
scores for assessment, is insufficiently sensitive to detect
small changes in MF which even so can serve as an import-
ant predictor of the effectiveness of the treatment method.
As the question of the effectiveness of BCI + exoskeleton
rehabilitation methods can be regarded as open [Teo and
Chew, 2014], its assessment needs methods excluding the
subjective component and giving numerical, physiological-
ly interpretable data on the state of MF to be used.
Recording of limb movements with subsequent biome-
chanical analysis of the parameters of these movements
provides such a method. This method is widely used to as-
sess the state of MF in poststroke and posttrauma patients
[Alt Murphy and Häger, 2015]. However, only two studies
have used this method for assessment of the state of MF
during rehabilitation using BCI + exoskeleton technology.
The first [Biryukova et al., 2016] analyzed the trajectory of
the working point of the paretic arm in movements to a sta-
tionary target, along with the contributions of the various
arm joints to these movements. Approximation of the speed
profile of the working point to a bell-shaped curve and inhi-
bition of joint movements not taking the working point to
the target were interpreted as improvements in overall con-
trol quality. In the second study [Kondur et al., 2016], angu-
lar accelerations at the arm joints were taken as assessments
of the total muscle torque and the number of local peaks in
angular speed as assessments of muscle spasticity driving
joint movement. After 10 BCI + exoskeleton procedures, a
patient with a duration of stroke of 4 months and severe
impairment to MF showed increases in muscle strength and
decreases in spasticity. Assessment of strength on the MRC
scale and spasticity on the MAS scale did not identify any
changes, while the overall assessment of MF on the Fugl-
Meyer scale increased. Each of these studies presented a
single clinical case of severe impairment to MF and report-
ed detailed biomechanical analysis of movements in all the
degrees of freedom of the arm. The results showed that this
analysis captures subclinical changes in movement parame-
ters. These changes can be regarded as predictors of suc-
cessful restoration of MF at the early stage of treatment and
can provide appropriate assessments of the potential of re-
habilitation using this method.
3. Plastic rearrangements after BCI + exoskeleton
procedures. Changes in MF seen during rehabilitation us-
ing BCI + exoskeleton technology are due to plastic rear-
rangements occurring in the cerebral cortex as a result of
training to kinesthetic motor imagery and its reinforcement
by proprioceptive feedback from movement of the exoskel-
Frolov, Kozlovskaya, Biryukova, and Bobrov
eton. Even cases of extensive cortical injury showed the
ability to exert voluntary control of the μ rhythm both in the
hemisphere ipsilateral to the injury and in the contralateral
hemisphere [Buch et al., 2008]. fMRI data showed stronger
activation in the contralateral hemisphere, as well as activa-
tion in its primary motor cortex, in a patient with hand ple-
gia who underwent a prolonged course of training to motor
imagery [Biryukova et al., 2016].
Analysis of fMRI data in a controlled study [Ramos-
Murguialday et al., 2013] showed that training to active
motor imagery led to a shift in activity in the motor and pre-
motor areas of the cortex from the contralateral (preserved)
hemisphere to the ipsilateral, i.e., to the side showing ac-
tivity in health. The index of lateralization of this activity
[Stinear et al., 2010] correlated with assessment of MF on
the Fugl-Mayer scale.
The study reported by Ono et al. [2014], conversely,
showed no correlation between improvements in MF and
the extent of desynchronization of the μ rhythm: activity in-
creased in both the study group and the control group, while
improvements in MF were seen only in the study group.
As regards the link between success in recognizing EEG
activity associated with motor imagery and improvement in
MF, these studies are contradictory: Ang et al. [2011] found
no correlation between them, while Ang et al. [2015] report-
ed a negative correlation and Frolov et al. [2016] found a
significant positive correlation. Differences in these results
may be associated with the different types of classifiers and
methods of evaluating MF used in these studies.
Ono et al. [2014] made the following suggestion be-
cause improvements in MF were seen only in the group us-
ing the exoskeleton. Desynchronization of the μ rhythm
could reflect arousal of the primary motor cortex during ex-
ecution of a significant motor task [Takemi et al., 2013].
Thus, the sensory input from the exoskeleton involving the
finger in motion may aid neuronal reorganization processes
in this region. Furthermore, because of the nonlinear nature
of neuron activity, additional arousal of the motor cortex by
somatosensory feedback may promote additional arousal of
the primary motor cortex, in turn helping to restore MF.
Conclusions. Despite use of intense rehabilitation
measures, restoration of the motor functions of the arms
for everyday self-care remains unachievable for most post-
stroke patients. Further improvements in and standardiza-
tion of rehabilitation procedures over recent years have led
to a multiplicity of procedures using robotic systems aiding
motor training in patients. However, these have not been as
encouraging as expected [Lo et al., 2010].
Rehabilitation effectiveness appears to depend not so
much on assistance to the arm – unweighting of the arm, the
type and number of movements performed by them – as the
extent of involvement of the patient in the treatment pro-
cess: concentration of attention and muscular and/or mental
efforts [Patton et al., 2008]. Passive robot therapy can even
improve recovery of MF and aid general relaxation of the
Use of Robotic Devices in Post-Stroke Rehabilitation
patient [Marchal-Crespo and Reinkensmeyer et al., 2009].
BCI controlling exoskeletons support the active involve-
ment of the patient in rehabilitation, even if the patient’s
motor resource is significantly restricted. Clinical studies of
BCI + exoskeleton procedures including a wide spectrum of
patient parameters – age, duration of stroke, extent of im-
pairment to MF – have demonstrated the validity of this
method (Table 4).
It should note noted, however, that specific studies of
the clinical effectiveness of this technology at different points
in recovery have not yet been reported and the retention of
progress in MF after procedures has not been studied, and
nor has the effect of repeated training. An open question is
that of the development of a standard rehabilitation protocol:
there are insufficient data to identify the optimum duration of
procedures or their intensity [Maciejasz et al., 2014].
The effectiveness of procedures has generally received
little study, though this depends on the type and severity
of damage. In these conditions, the task of identifying tar-
get groups of patients with rehabilitation potential in whom
BCI + exoskeleton treatment will be maximally effective is
critical [Kotov and Stakhovskaya, 2014]. As noted above,
this treatment is indicated in patients with severe pareses
and plegias as the only active type of rehabilitation avail-
able. Our data [Frolov et al., 2016] indicate that the target
group can also include patients with moderate pareses at
different stages of rehabilitation. Training to motor imagery
can aid the generalization of spontaneous recovery mecha-
nisms, which operate most actively in the first three months
after stroke [Zeiler and Krakauer, 2013] and can stimulate
learning mechanisms in preserved brain structures during
subsequent periods of rehabilitation. An important condi-
tion in selecting patients is preservation of cognitive ability,
the ability to maintain concentration for long periods, and
motivation to take part in treatment [Chernikova, 2016].
Successful learning to control an exoskeleton with BCI
based on kinesthetic motor imagery leads to recovery of the
integrative activity of the brain, giving hope for improve-
ment not only of motor, but also of cognitive functions.
Acknowledgements. This work was supported by the
Russian Foundation for Basic Research (Grant Nos. 16-29-
08247 ofi_m, 16-29-08206 ofi_m, 16-04-01506, and 16-04-
00962.
Abbreviations:
MF – motor function;
BCI – brain–computer interface;
fMRI – functional magnetic resonance imaging;
CNS – central nervous system;
EMG – electromyogram;
EEG – electroencephalogram.
REFERENCES
Alexandrov, A. V. and Frolov, A. A., “Closed-loop and open-loop control
of posture and movement during human upper trunk bending,” Biol.
Cybern., 104, No. 6, 425–438 (2011).
1063
Alexandrov, A. V., Frolov, A. A., Horak, F. B., Carlson-Kuhta, P., and
Park, S., “Feedback equilibrium control during human standing,”
Biol. Cybern., 93, 309–322 (2005).
Alt Murphy, M. A. and Häger, C. K., “Kinematic analysis of the upper ex-
tremity alter stroke – how far have we reached and what have we
grasped?” Phys. Ther. Rev., 20, 137–155 (2015).
Aman, K. and Al-Jumailyb, A. A., “Active exoskeleton control systems:
state of the art,” Procedia Engineering, 41, 988–994 (2012).
Amirabdollahian, F., Loureiro, R., Gradwell, E., Collin, C., Harwin, W.,
and Johnson, G., “Multivariate analysis of the Fugl-Meyer outcome
measures assessing the effectiveness of GENTLE/S robot-mediated
stroke therapy,” J. Neuroeng. Rehabil., 4, No. 1, 4 (2007).
Ang, K. K., Chua, K. S., Phua, K. S., Wang, C., Chin, Z. Y., Kuah, C. W.,
Low, W., and Guan, C., “A randomized controlled trial of EEG-
based motor imagery brain–computer interface robotic rehabilitation
for stroke,” Clin. EEG Neurosci., 46, No. 4, 310–320 (2015).
Ang, K. K., Guan, C., Chua, K. S., and Ang, B. T., Kuah, C. W., Wang, C.,
Phua, K. S., Chin, Z. Y., and, Zhang, H., “A large clinical study on
the ability of stroke patients to use an EEG-based motor imagery
brain–computer interface,” Clin. EEG Neurosci., 42, No. 4, 253–258
(2011).
Ang, K. K., Guan, C., Phua, K. S., Wang, C., Zhou, L., Tang, K. Y.,
Ephraim Joseph, G. J., Kuah, C. W., “Brain–computer interface-
based robotic end effector system for wrist and hand rehabilitation:
results of a three-armed randomized controlled trial for chronic
stroke,” Front. Neuroeng., 7, 30 (2014).
Bach-Y-Rita, P., “Theoretical and practical considerations in the resto-
ration of function after stroke,” Top. Stroke Rehabil., 8, No. 3, 1–15
(2001).
Balasubramanian, S., Wei, R., Perez, M., Shepard, B., Koeneman, E., Koe-
neman, J., and He, J., “RUPERT: An exoskeleton robot for assisting
rehabilitation of arm functions,” Proceedings of the Conference
Virtual Rehabilitation, Vancouver, Canada (2008), pp. 163–167.
Bernshtein, N. A., Dexterity and its Development, Fizkul’tura i Sport,
Moscow (1947/1991).
Bernshtein, N. A., Studies of the Biodynamics of Locomotion, VIEM,
Moscow, Leningrad (1935).
Bernshtein, N. A., The Construction of Movements, Meditsina, Moscow
(1947).
Biryukova, E. V., Pavlova, O. G., Kurganskaya, M. E., Bobrov, P. D.,
Turbina, L. G., Frolov, A. A., Davydov, V. I., Sil’chenko, A. V., and
Mokienko, O. A., “Recovery of the motor function of the arm using
a hand exoskeleton controlled by a ‘brain–computer’ interface. A
case of a patient with extensive damage to brain structures,” Fiziol.
Cheloveka, 42, No. 1, 19–30 (2016).
Biryukova, E. V., Roschin, V. Y., Frolov, A. A., Ioffe, M. E., Massion, J.,
and Dufosse, M., “Forearm postural control during unloading: antic-
ipatory changes in elbow stiffness,” Exp. Brain Res., 124, No. 1,
107–117 (1999).
Bizzi, E., Acconero, N., Chapple, W., and Hogan, N., “Arm trajectory for-
mation,” Exp. Brain Res., 46, 139–143 (1982).
Bobrov, P. D., Isaev, M. R., Korshakov, A. V., Oganesyan, V. V., Kerecha-
nin, Ya. V., Popod’ko, A. I., and Frolov, A. A., “Sources of electro-
physiological and foci of hemodynamic activity in the brain signifi-
cant for the control of a hybrid brain–computer interface based on
the recognition of EEG patterns and near infrared spectrograms on
motor imagery,” Fiziol. Cheloveka, 42, No. 3, 12–24 (2016).
Bobrov, P., Frolov, A. A., and Húsek, D., “Brain computer interface en-
hancement by independent component analysis,” in: Proceedings of
the Third International Conference on Intelligent Human Computer
Interaction (IHCI 2011), Prague, Czech Republic, August, 2011,
Springer, Berlin, Heidelberg (2011), pp. 51–60.
Bos, R. A., Haarman, C. J. W., Stortelder, T., Nizamis, K., Herder, J. L.,
Stienen, A. H. A., and Plettenburg, D. H., “A structured over-
view of trends and technologies used in dynamic hand orthoses,”
J. Neuroeng. Rehabil., 13, 1 (2016).
1064
Brauchle, D., Vukelic, M., Bauer, R., and Gharabaghi, A., “Brain state-de-
pendent robotic reaching movement with a multi-joint arm exoskel-
eton: combining brain-machine interfacing and robotic rehabilita-
tion,” Front. Hum. Neurosci., 9, 564 (2015).
Brewer, B. R., Klatzky, R., and Matsuoka, Y., “Visual feedback distortion
in a robotic environment for hand rehabilitation,” Brain Res. Bull.,
75, No. 6, 804–813 (2008).
Buch, E., Weber, C., Cohen, L. G., Braun, C., Dimyan, M. A., Ard, T.,
Mellinger, J., Caria, A., et al., “Think to move: a neuromagnetic
brain–computer interface (BCI) system for chronic stroke,” Stroke,
39, 910–917 (2008).
Cordo, P., Lutsep, H., Cordo, L., Wright, W. G., Cacciatore, T., and Skoss, R.,
“Assisted movement with enhanced sensation (AMES): coupling mo-
tor and sensory to remediate motor deficits in chronic stroke patients,”
Neurorehabil. Neural Repair, 23, 67–77 (2009).
Delorme, A., Palmer, J., Onto, J., Oostenveld, R., and Makeig, S., “Indepe-
ndent EEG sources are dipolar,” PLoS One, 7, No. 2, e30135 (2012).
Di Pino, G., Pellegrino, G., Assenza, G., Capone, F., Ferreri, F., Formica, D.,
Ranieri, F., Tombini, M., “Modulation of brain plasticity in stroke:
a novel model for neurorehabilitation,” Nat. Dev. Neurol., 10, 597–
608 (2014).
Dipietro, L., Ferraro, M., Palazzolo, J. J., Krebs, H. I., Volpe, B. T., and
Hogan, N., “Customized interactive robotic treatment for stroke:
EMG triggered therapy,” IEEE Trans. Neur. Syst. Rehab. Eng., 13,
No. 3, 325–334 (2005).
Dovat, L., Lambercy, O., Gassert, R., Maeder, T., Milner, T., Leong, T. C.,
and Burdet, E., “HandCARE: a cable-actuated rehabilitation system
to train hand function after stroke,” IEEE Trans. Neural Syst.
Rehabil. Eng., 16, No. 6, 582–591 (2008).
Fazekas, G., Horwath, M., and Toth, A., “A novel robot training system
designed to supplement upper limb physiotherapy on patient with
spastic hemiparesis,” Int. J. Rehabil. Res., 29, 251–254 (2006).
Fel’dman, A. G., Central and Reflex Mechanisms of Motor Control, Nauka,
Moscow (1979).
Ferris, D. P., “The exoskeletons are here,” J. Neuroeng. Rehabil., 6, 17 (2009).
Frisoli, A., Borelli, L., Montagner, A., Marcheschi, S., Procopio, C.,
Salsedo, F., Bergamasco, M., Carboncini, M. C., et al., “Arm reha-
bilitation with a robotic exoskeleton in virtual reality,” Proceedings
of the IEEE 10th International Conference on Rehabilitation and
Robotics (ICORR), Noordwijk, Netherlands (2007), pp. 631–642.
Frolov, A. A., Dufosse, M., Rizek, S., and Kaladjan, A., “On the possibility
of linear modeling of the human arm neuromuscular apparatus,”
Biol. Cybern., 82, No. 6, 499–515 (2000).
Frolov, A. A., Gusek, D., Sil’chenko, A. V., Tintera, Ya., and Rydlo, Ya.,
“Changes in the hemodynamic activity of the brain in motor imagery
as a result of training subjects to control a brain–computer inter-
face,” Fiziol. Cheloveka, 42, No. 1, 5–18 (2016a).
Frolov, A. A., Husek, D., Biryukova, E., V., Bobrov, P. D., Mokienko, O. A.,
and Alexandrov, A. V., “Principles of motor recovery in post-stroke
patients using hand exoskeleton controlled by the brain–computer
interface based on motor imagery,” Neural Network World, 27, No. 1,
107–137 (2017).
Frolov, A. A., Mokienko, O. A., Lyukmanov, R., Chernikova, L. A., Ko-
tov, S. V., Turbina, L. G., Bobrov, L. D., Biryukova, E. V., et al.,
“Preliminary results of a controlled study of the effectiveness of
BCI-exoskeleton technology in poststroke paresis of the arm,” Vestn.
Ros. Gos. Med. Univ., 2, 17–25 (2016).
Frolov, A. A., Prokopenko, R. A., Dufosse, M., and Ouezdou, F. B.,
“Adjustment of the human arm viscoelastic properties to the direc-
tion of reaching,” Biol. Cybern., 94, 97–109 (2006).
Frolov, A. A., Roschin, V. Y., and Biryukova, E. V., “Adaptive neural net-
work model of multijoint movement control by working point veloc-
ity,” Neural Network World, 4, No. 2, 141–156 (1994).
Frolov, A., Húsek, D., Bobrov, P., Mokienko, O., and Tintera, J., “Sources
of electrical brain activity most relevant to performance of brain–
computer interface based on motor imagery,” in: Brain–Computer
Frolov, Kozlovskaya, Biryukova, and Bobrov
Interface Systems – Recent Progress and Future Prospects (2013),
pp. 175–193.
Frolov, A., Misek, D., Bobrov, P., Korshakov, A., Chernikova, L.,
Konovalov, R., and Mokienko, O., “Sources of EEG activity most
relevant to performance of brain–computer interface based on motor
imagery,” Neural Network World, 22, No. 1, 21–37 (2012).
Fugl-Meyer, A. R., Jääskö, L., Leyman, L., Olsson, S., and Stegling, S.,
“The poststroke hemiplegic patient. A method for evaluation of
physical performance,” Scand. J. Rehabil. Med., 7, 13–31 (1975).
Gladstone, D. J., Daniells, C. J., and Black, S. E., “The Fugl-Meyer
Assessment of motor recovery alter stroke: A critical review of its
measurement properties,” Neurorehabil. Neural Repair, 16, 232–
240 (2002).
Gomi, H. and Kawato, M., “Equilibrium-point control hypothesis exam-
ined by measured arm stiffness during multijoint movement,”
Science, 272, 117–120 (1996).
Grèzes, J. and Decety, J., “Functional anatomy of execution, mental simu-
lation, observation verb generation of actions: a meta-analysis,”
Hum. Brain Mapp., 12, No. 1, 1–19 (2001).
Grosse-Wentrup, M., Mattia, D., and Oweiss, K., “Using brain–computer
interfaces to induce neural plasticity and restore function,” J. Neural
Eng., 8, No. 2, 025004 (2011).
Heo, P., Gu, G. M., Lee, S., Rhee, K., and Kim, J., “Current hand exoskel-
eton technologies for rehabilitation and assistive engineering,” Int. J.
Precision Eng. Manufact., 13, No. 5, 807–824 (2012).
Hesse, S., Kuhlmann, H., Wilk, J., Tomelleri, C., and Kirker, S. G. B.,
“A new electromechanical trainer for sensorimotor rehabilitation of
paralysed fingers: a case series in chronic and acute stroke patients,”
J. Neuroeng. Rehabil., 5, 21 (2008).
Hesse, S., Schulte-Tigges, G., Konrad, M., Bardeleben, A., and Werner, C.,
“Robot-assisted arm trainer for the passive and active practice of
bilateral forearm and wrist movements in hemiparetic subjects,”
Arch. Phys. Med. Rehabil., 84, No. 6, 915–920 (2003).
Hétu, S., Gregoire, M., Saimpont, A., Coll, M. P., Eugène, F., Michon, P. E.,
and Jackson, P. L., “The neural network of motor imagery: An ALE
meta-analysis,” Neurosci. Biobehav. Rev., 37, 930–949 (2013).
Ho, N. S. K., Tong, K. Y. Hu, X. L., Fung, K. L., Wei, X. J., Rong, W., and
Susanto, E. A., “An EMG-driven exoskeleton hand robotic training
device on chronic stroke subjects: task training system for stroke
rehabilitation,” IEEE Int. Conf. Rehabil. Robot, Boston, MA (2011),
2011:5975340.
Hogan, N. and Flash, T., “Moving gracefully: quantitative theories of mo-
tor coordination,” Trends Neurosci., 10, 170–174 (1987).
Jack, D., Boian, R., Merlans, A. S., Tremaine, M., Burdea, G. C., Adamo-
vich, S. V., Recce, M., and Poizner, H., “Virtual reality-enhanced
stroke rehabilitation,” Neural Syst. Rehabil. Eng. IEEE Trans., 9,
No. 3, 308–318 (2001).
Jeannerod, M. and Frak, V., “Mental imaging of motor activity in humans,”
Curr. Opin. Neurobiol., 9, 735–739 (1999).
Jeannerod, M., “The representing brain: Neural correlates of motor inten-
tion and imagery,” Behav. and Brain Sci, 17, No. 2, 187–202 (1994).
Kachenoura, A., Albera, L., Senhadji, L., and Common, P., “ICA: a poten-
tial tool for BCI systems,” IEEE Signal Process. Mag., 25, No. 1,
57–68 (2008).
Kaneko, Y., Nakano, E., Osu, R., Wada, Y., and Kawato, M., “Trajectory
formation based on the minimum commanded torque change model
using the Euler-Poisson equation,” Systems and Computers in Japan,
36, 92–103 (2005).
Klimesch, W., “Alpha-band oscillations, attention-controlled access to
stored information,” Trends Cogn. Sci., 16, No. 12, 606–617 (2012).
Klimesch, W., Sauseng, P., and Hanslmayr, S., “EEG alpha oscillations: the
inhibition-timing hypothesis,” Brain Res. Rev., 53, No. 1, 63–88
(2007).
Koeneman, E. J., Schultz, R. S., Wolf, S. L., Herring, D. E., and Koeneman,
J. B., “A pneumatic muscle hand therapy device,” Conf. Proc. IEEE
Eng. Med. Biol. Soc., 4, 2711–2713 (2004).
Use of Robotic Devices in Post-Stroke Rehabilitation
Kondur, A. A., Biryukova, E. V., Kotov, S. V., Turbina, L. G., and Fro-
lov, A. A., “Kinematic portrait of the patient as an objective measure
of the state of motor function during the process of neurorehabilita-
tion using an arm exoskeleton controlled by a brain–computer inter-
face,” Uchen. Zapiski, St. Petersb. Med. Univ. im. I. P. Pavlova, 23,
No. 3, 28–31 (2016).
Kotov, S. V. and Stakhovskaya, L. V., Stroke, MIA, Moscow (2014).
Kotov, S. V., Turbina, L. G., Bobrov, P. D., Frolov, A. A., Pavlova, O. G.,
Kurganskaya, M. E., and Biryukova, E. V., “Rehabilitation of stroke
patients using a bioengineered ‘brain–computer interface + exoskel-
eton’ system,” Zh. Nevrol. Psikhiatr. im. S. S. Korsakova, 12, 66–72
(2014).
Krebs, H. I., Hogan, N., Aisen, M. L., and Volpe, B. T., “Robot-aided neu-
rorehabilitation,” IEEE Trans. Rehabil. Eng., 6, 75–87 (1988).
Kwakkel, G. and Meskers, C. G. M., “Effects of robotic therapy of the arm
after stroke,” Lancet Neurol., 13, No. 2, 132–133 (2014).
Leont’ev, A. N. and Zaporozhets, A. V., The Recovery of Movements,
Sovetskaya Nauka, Moscow (1945).
Levin, M. F., “Interjoint coordination during pointing movements is dis-
rupted in spastic hemiparesis,” Brain, 119, 281–294 (1996).
Lo, A. C., Guarino, P. D., Richards, L. G., Haselkorn, J. K., Wittenberg, G. F.,
Federman, D. G., Ringer, R. J., Wagner, T. H., et al., “Robot-assisted
therapy for long-term upper-limb impairment after stroke,” New Engl.
J. Med., 362, 1772–1783 (2010).
Lotze, M., Braun, C., Birbaumer, N., Anders, S., and Cohen, L. G., “Motor
learning elicited by voluntary drive,” Brain, 126, No. 4, 866–872
(2003).
Lum, P. S., Burgar, C. G., Van der Loos, M., Shor, P. C., Majmundar, M.,
and Yap, R., “MIME robotic device for upper-limb neurorehabilita-
tion in subacute stroke subjects: A follow-up study,” J. Rehabil. Res.
Dev., 43, No. 5, 631–642 (2006).
Maciejasz, P., Eschweiler, J., Gerlach-Hahn, K., Jansen-Troy, A., and
Leonhardt, S., “A survey on robotic devices for upper limb rehabili-
tation,” J. NeuroEng. Rehabil., 11, 3 (2014).
Makeig, S., Bell, A. J., Jung, T. P., and Sejnowski, T. J., “Independent com-
ponent analysis of electroencephalographic data,” Adv. Neural Info.
Proc. Syst., 8, 145–151 (1996).
Marchal-Crespo, L. and Reinkensmeyer, D. J., “Review of control strategies
for robotic movement training after neurologic injury,” J. Neuroeng.
Rehabil., 6, 20 (2009).
Masia, L., Krebs, H. I., Cappa, P., and Hogan, N., “Design, characteriza-
tion, and impedance limits of a hand robot,” Proc. IEEE 10th Int.
Conf. Rehabil. Robotics, (ICORR), Noordwijk, Netherlands (2007),
pp. 1085–1089.
Mayr, A., Kofler, M., and Saltuari, L., “ARMOR: an electromechanical
robot for upper limb training following stroke. A prospective ran-
domized controlled pilot study,” Handchir. Mikrochir. Plast. Chir.,
40, 66–73 (2008).
Micera, S., Carrozza, M., Guglielmelli, E., Cappiello, G., Zaccone, F.,
Freschi, C., Colombo, R., Mazzone, A., et al., “A simple robotic sys-
tem for neurorehabilitation,” Autonomous Robots, 19, No. 3, 271–
284 (2005).
Mokienko, O. A., Bobrov, P. D., Chernikova, L. A., and Frolov, A. A., “A
motor imagery-based brain–computer interface in the rehabilitation
of patients with hemiparesis,” Byull. Sibirsk. Med., 12, No. 2, 30–35
(2013).
Mokienko, O. A., Lyukmanov, R. Kh., Chernikova, L. A., Suponeva, N. A.,
Paradov, M. A., and Frolov, A. A., “A brain–computer interface: first
experience in clinical use in Russia,” Fiziol. Cheloveka, 42, No. 1,
31–39 (2016).
Mulas, M., Folgheraiter, M., and Gini, G., “An EMG-controlled exoskele-
ton for hand rehabilitation,” Proc. 9th Int. Conf. Rehabil. Robotics
ICORR, Chicago, IL, (2005), pp. 371–374.
Nef, T., Guidaili, M., Klamroth-Marganska, V., and Riener, R., “ARMin –
exoskeleton robot for stroke rehabilitation,” World Congress on
Medical Physics and Biomedical Engineering, September 7–12,
1065
IFMBE Proceedings, Dössel, O. and Schlegel, W. C. (eds.), Springer,
Munich, Germany (2009), pp. 127–130.
Nudo, R. J., Milliken, G. W., Jenkins, W. M., and Merzenich, M. M., “Use-
dependent alterations of movement representations in primary motor
cortex of adult squirrel monkeys,” J. Neurosci., 16, No. 2, 785–807
(1996).
Ono, T., Shindo, K., Kawashima, K., Ota, N., Ito, M., Ota, T., Mukaino, M.,
Fujiwara, T., Kimura, A., Liu, M., et al., “Brain–computer interface
with somatosensory feedback improves functional recovery from se-
vere hemiplegia due to chronic stroke,” Front. Neuroeng., 7, 19 (2014).
Onton, J., Westerfield, M., Townsend, J., and Makeig, S., “Imaging human
EEG dynamics using independent component analysis,” Neurosci.
Biobehav. Rev., 30, No. 6, 808–822 (2006).
Patton, J. L., Small, S. L., and Rymer, W. Z., “Functional restoration for the
stroke survivor: informing the efforts of engineers,” Top Stroke
Rehabil., 15, No. 6, 521–541 (2008).
Patton, J. L., Stoykov, M. E., Kovic, M., and Mussa-Ivaldi, F., “Evaluation
of robotic training forces that either enhance or reduce error in
chronic hemiparetic stroke survivors,” Exp. Brain Res., 168, No. 3,
368–383 (2006).
Pedrocchi, A., Ferrante, S., Ambrosini, E., Gandolla, M., Casellato, C.,
Schauer, T., Klauer, C., Pascual, J., et al., “MUNDUS project: Multi-
modal neuroprosthesis for daily upper limb support,” J. Neuroeng.
Rehabil., 10, 66 (2013).
Peterka, R. J., “Sensorimotor integration in human postural control,”
J. Neurophysiol., 88, 1097–1118 (2002).
Pfurtscheller, G. and Berghold, A., “Patterns of cortical activation during
planning of voluntary movement,” Electroencephalogr. Clin. Neuro-
physiol., 72, 250–258 (1989).
Pfurtscheller, G. and Lopes da Silva, F. H., “Event-related EEG/MEG syn-
chronization and desynchronization: basic principles,” Clin. Neuro-
physiol., 110, 1842–1857 (1999).
Pfurtscheller, G., “EEG event-related desynchronization (ERD) and event
related synchronization (ERS),” in: Electroencephalography: Basic
Principles, Clinical Applications and Related Fields, Niedermeyer,
E. and Lopes da Silva, F. H., (eds.) Williams and Wilkins, Baltimore,
MD (1999), 4th edition, pp. 958–967.
Pignolo, L., Dolce, G., Basta, G., Lucca, L. F., Serra, S., and Sannitam W. G.,
“Upper limb rehabilitation after stroke: ARAMIS a ‘robo-mechatron-
ic’ innovative approach and prototype,” 4th IEEE RAS & EMBS Int.
Conf. Biomedical Robotics and Biomechatronics (BioRob), Rome,
Italy (2012), pp. 1410–1414.
Ramos-Murguialday, A., Broetz, D., Rea, M., Laer, L., Yilmaz, O., Brasil,
F. L., Liberati, G., Curado, M. R., et al., “Brain-machine interface in
chronic stroke rehabilitation: a controlled study,” Ann. Neurol., 74,
No. 1, 100–108 (2013).
Reinkensmeyer, D. J., Kahn, L. E., Averbuch, M., McKenna-Cole, A. N.,
Schmit, B. D., and Rymer, W. Z., “Understanding and treating arm
movement impairment after chronic brain injury: Progress with the
ARM Guides,” J. Rehabil. Res. Dev., 37, No. 6, 653–662 (2000).
Ren, Y., Park, H. S., and Zhang, L. Q., “Developing a whole-arm exoskel-
eton robot with hand opening and closing mechanism for upper limb
stroke rehabilitation,” Proc. IEEE Int. Conf. Rehabil. Robotics
(ICORR), Kyoto, Japan (2009), pp. 761–765.
Restorative Neurology: Innovatory Technologies in Neurorehabilitation,
Chernikova, L. A. (ed.), Medical Information Agency Press, Moscow
(2016).
Riener, R., Nef, T., and Colombo, G., “Robot-aided neurorehabilitation of
the upper extremities,” Med. Biol. Eng. Comput., 43, 2–10 (2005).
Rosati, G., Gallina, P., and Masiero, S., “Design, implementation and clin-
ical tests of a wire-based robot for neurorehabilitation,” IEEE Trans.
Neural Syst. Rehabil. Eng., 15, No. 4, 560–569 (2007).
Sanchez, R., Reinkensmeyer, D., Shah, P., Liu, J., Rao, S., Smith, R., Cramer,
S., Rahman, T., and Bobrow, J., “Monitoring functional arm move-
ment for home-based therapy after stroke,” Conf. Proc. IEEE Eng.
Med. Biol. Soc. San Francisco, CA (2004), Vol. 7, pp. 4787–4790.
1066
Schabowsky, C. N., Godfrey, S. B., Holley, R. J., and Lum, P. S., “Develop-
ment and pilot testing of HEXORR: hand EXOskeleton rehabilita-
tion robot,” J. Neuroeng. Rehabil., 7, 36 (2010).
Shaw, S. E., Morris, D. M., Uswatte, G., McKay, S., Meythaler, J. M., and
Taub, E., “Constraint-induced movement therapy for recovery of up-
per-limb function following traumatic brain injury,” J. Rehabil. Res.
Dev, 42, No. 6, 769–778 (2005).
Shelton, F. N. A. P. and Reding, M. J., “Effect of lesion location on upper
limb motor recovery after stroke,” Stroke, 32, 107–112 (2001).
Shih, J. J., Krusienski, D. J., and Wolpaw, J. R., “Brain–computer interfac-
es in medicine,” Mayo Clin. Proc., 87, No. 3, 268–279 (2012).
Sinel’nikov, R. D., Atlas of Human Anatomy, Meditsina, Moscow (1967),
Vol. 1.
Soekadar, S. R., Birbaumer, N., Slutzky, M. W., and Cohen, L. G., “Brain-
machine interfaces in neurorehabilitation of stroke,” Neurobiol. Dis.,
83, 172–179 (2015).
Stienen, A., Hekman, E., Prange, G., Jannink, M., Aalsma, A., van der
Helm, F., van der Kooij, H., “Dampace: Design of an exoskeleton for
force-coordination training in upper extremity rehabilitation,” J. Med.
Devices, 3, 031003 (2009).
Stinear, C. M., “Prediction of recovery of motor function after stroke,”
Lancet Neurol., 9, No. 12, 1228–1232 (2010).
Sun, H., Blakely, T. M., Darvas, F., Wander, J. D., Johnson, L. A., Su, D. K.,
Miller, K. J., Fetz, E. E., et al., “Sequential activation of premotor,
primary somatosensory and primary motor areas in humans during
cued finger movements,” Clin. Neurophysiol., 126, No. 11, 2150–
2161 (2015).
Frolov, Kozlovskaya, Biryukova, and Bobrov
Takahashi, C. D., Der-Yeghiaian, L., Le, V., Motiwala, R. R., and Cra-
mer, S. C., “Robot-based hand motor therapy after stroke,” Brain,
131, 425–437 (2008).
Takemi, M., Masakado, Y., Liu, M., and Ushiba, J., “Event-related desyn-
chronization reflects down-regulation of intracortical inhibition in hu-
man primary motor cortex,” J. Neurophysiol., 110, 1158–1166 (2013).
Taub, E., Uswatte, G., and Elbert, T., “New treatments in neurorehabilita-
tion founded on basic research,” Nat. Rev. Neurosci., 3, No. 3, 228–
236 (2002).
Teo, W. P. and Chew, E., “Is motor imagery brain–computer interface fea-
sible in stroke rehabilitation?,” PM R, 6, 723–728 (2014).
Tong, K. Y., Ho, S. K., Pang, P. K., Hu, X. L., Tam, W. K., Fung, K. L.,
Wei, X. J., Chen, P. N., et al., “An intention driven hand functions
task training robotic system,” Conf. Proc. IEEE Eng. Med. Biol. Soc.
V. Buenos Aires, Argentina (2010), pp. 3406–3409.
Varkuti, B., Guan, C., Pan, Y., Phua, K. S., Ang, K. K., Kuah, C. W. K.,
Chua, K., Ang, B., T., et al., “Resting state changes in functional
connectivity correlate with movement recovery for BCI and ro-
bot-assisted upper-extremity training after stroke,” Neurorehabil.
Neural Repair, 27, 53–62 (2013).
Wolbrecht, E. T., Leavitt, J., Reinkensmeyer, D. J., and Bobrow, J. E.,
“Control of a pneumatic orthosis for upper extremity stroke rehabili-
tation,” Conf. Proc. IEEE Eng. Med. Biol. Soc., 1, 2687–2693 (2006).
Wright, Z. A., Rymer, W. Z., and Slutzky, M. W., “Reducing abnormal mus-
cle coactivation after stroke using a myoelectric computer interface:
A pilot study,” Neurorehabil. Neural Repair, 28, 443–451 (2014).
Zeiler, S. R. and Krakauer, J. W., “The interaction between training and
plasticity in the post-stroke brain,” Curr. Opin. Neurol., 26, No. 6,
609–616 (2013).