Is robot-aided sensorimotor training in stroke rehabilitation a
realistic option?
Bruce T. Volpea,b, Hermano I. Krebsc and Neville Hoganc,d
Stroke is the leading cause of disability, despite continued
advances in prevention and treatment techniques based on
novel delivery of new fibrinolytic drugs. Improved medical
treatment of the complications caused by acute stroke has
contributed to decreased mortality, but 90% of the survivors
have significant neurological deficits. Reducing the degree of
permanent disability remains the goal of poststroke neurorehabilitation programs, and new approaches to impairment
reduction through managing sensorimotor experience may
contribute further to altering disability. Recent reports from a
number of laboratories using enhanced sensorimotor training
protocols, particularly those with robotic devices, have indicated
modest success in reducing impairment and increasing motor
power in the exercised limb of patients with stroke when
compared with control individuals. Whether arming the therapist
with new tools, especially robotic devices, to treat impairment is
a realistic approach to modern interdisciplinary rehabilitation
raises questions regarding the added value of impairment
reduction, and under what conditions should scientific and
clinical development of robotic studies continue. Curr Opin Neurol
14:745±752.
#
2001 Lippincott Williams & Wilkins.
a
Burke Medical Research Institute, White Plains, New York, bWeill Medical College of
Cornell University, Department Neurology and Neuroscience, NY, NY,
c
Massachusetts Institute of Technology, Mechanical Engineering Department,
Newman Laboratory for Biomechanics and Human Rehabilitation, Massachusetts,
and dMassachusetts Institute of Technology, Brain and Cognitive Sciences
Department, Cambridge, MA, USA
Correspondence to Bruce T. Volpe, Burke Medical Research Institute,
785 Mamaroneck Avenue, White Plains, NY 10605, USA
Fax: +1 914 597 2796; e-mail: bvolpe@burke.org
Current Opinion in Neurology 2001, 14:745±752
# 2001 Lippincott Williams & Wilkins
1350-7540
Introduction
Despite the success of preventive strategies that control
blood pressure and treat atrial ®brillation, and that
encourage cessation of smoking, attention to weight
control and a modest exercise regimen, recent studies
have revised upward the incidence of stroke [1,2]. With
increasing life expectancy and the coincidence of the
`baby boom' generation growing older, coupled with
improved medical treatment of the complications caused
by acute stroke, in the USA the ranks of the over 4
million survivors of stroke alive today are likely to swell
considerably [1]. Most survivors of stroke are left with
signi®cant residual physical, cognitive and psychological
impairments, and the combination of these impairments
results in disability.
Re-emergence of the treatment principle
of impairment reduction
A patient's impairment or neurologic de®cit is de®ned by
the speci®c loss or abnormality of psychologic, physiologic, or anatomic structure. Impairment after stroke
depends on the lesion size and location in the brain.
Disability is a broader term that captures any restriction
or lack of ability to perform an activity within the range
considered normal. Current interdisciplinary treatment
programs for patients with stroke attend to both
impairment and disability reduction; however, the rush
to discharge the patient from the rehabilitation hospital
more quickly has prompted a shift toward encouraging
functional improvement by learning compensatory techniques. Although this emphasis on disability reduction
also happens to coincide with the patient's main
concern, it occurs at the expense of impairment
reduction.
Recent experimental results [3] suggest that the hasty
compensation for a disability engenders a pattern of
disuse in the impaired limbs that extinguishes temporarily that aspect of recovery and mutes the potential
for future impairment change, if not recovery measured
in terms of disability reduction. In fact, constraint
induced or forced-use treatments, in which the partly
paralyzed upper limb is the focus of motor activity,
have led to signi®cant improvements in motor power
both in the acute [4] and chronic [5,6] rehabilitation
period after stroke. Studies of patients on acute
poststroke rehabilitation programs had the advantage
of comparing a forced-use treated group with control
individuals [4].
745
746 Trauma and rehabilitation
Other studies [5,6] using constraint-induced programs in
patients with chronic stroke suggest that, for some,
additional recovery follows additional training. The
largest proportion of recovery occurs during the weeks
and months that immediately follow the acute stroke;
these data therefore suggest that, for some, motor
recovery occurs in a salutatory pattern, with functional
plateaus that may be followed by change. Rehabilitation
and attention to recovery need not stop after the acute
rehabilitation hospital event, and the potential of home
training or home training enhanced with devices
managed by therapists might continue to contribute to
recovery goals.
Rigorous outcome studies with appropriate
controls trigger the need for new approaches
Forced-use or constraint-induced strategies have
exploited the natural context of patients' environment
by requiring increased use of the paralyzed limb.
These efforts are consistent with experiments in
which increased focused additional sensorimotor training for the paralyzed limb were employed to improve
outcome [7,8]. Because most patients display some
recovery after stroke and the natural history of those
changes (especially for the sensorimotor system) have
been well described, recent studies of the effect of a
speci®c intervention have compared outcome with an
appropriate control. For example, investigators have
demonstrated that the addition of 30 min of a pushing
exercise of the paretic upper limb over 30 sessions to
a program of poststroke rehabilitation facilitated motor
recovery of that paretic limb [9]. A general attempt to
enhance rehabilitation with an ``eclectic . . . selection
of treatment techniques'' led to improved motor
outcome 6 months after the stroke, but the control
group had caught up with the treatment group by 1
year [10,11]. Other experiments also attempted to
focus the training and then specify the measurement
of outcome [7]. In that study, enhanced training for
the upper or lower limb led to improved outcome for
treated limbs, and not especially for the `untreated'
limbs.
Several groups have described results from functional
imaging of the brain in patients recovered from stroke
[12±16]. There was increased blood ¯ow in areas around
the lesion, in supplemental and premotor cortex, and in
ipsilateral motor cortex (ipsilesional). The signi®cance of
these novel descriptions remains controversial. In
patients recovering from stroke, recent work has tested
the effect of enhanced treatment for the paralyzed arm
using sequential positron emission tomography images
performed while the patients had their affected limb
passively moved [17 .]. The second positron emission
tomography scan, in patients randomized to an intensive
treatment or standard treatment, demonstrated that the
enhanced treatment group had more activation, indicating greater regional cerebral blood ¯ow in a number of
sensorimotor cortical areas. The group treated with
enhanced therapy also had improved functional outcome
as compared with control individuals on standard
treatment. Increased regional activation was also observed in a functional magnetic resonance imaging task
in patients recovering from stroke [18 .]. In those
controlled studies the group treated with `more' therapy
had better motor outcome, particularly when the outcome measure was focused on the exercised limb.
Although the effect on muscles, joints, and the general
effect of conditioning were not measured in those
studies, the signi®cant changes in regional cerebral
blood ¯ow suggest that enriched sensorimotor experience has a direct effect on the brain.
A broader context for candidate brain
mechanisms underlying recovery
Recent work with animal recovery models also supports
the idea that training enhances recovery after damage to
the central nervous system. Animals with focal cortical
injury exposed to enriched or challenging sensorimotor
environments exhibited greater anatomic responses [19±
21]. Other experiments in animals in which highly
practiced motor tasks were interrupted by speci®c focal
brain injury [22±26] demonstrated that retraining the
motor-impaired animals led to improved functional
output, sometimes nearly to levels of prelesion performance, depending on the task and lesion size. The
critical measures focused on remapping the sensory
cortex in animals that had sustained lesions, had been
trained, and in which the prelesion maps had been
determined. Postlesion sensory representations of the
trained upper limb re-emerged in novel cortical locations. Furthermore, the studied cortical zones were
excited by novel stimuli, the representations were
enlarged, and multiple receptive ®elds from other parts
of the limb were also represented in the emergent
cortical receptive ®eld.
Basic work in animal models, clinical outcomes research,
and functional brain imaging data suggest that activitydependent plasticity contributes importantly to recovery.
This work supports a clinical model in which some
sensorimotor experience abets recovery of the impairment after brain injury.
Present state of therapy: an opportunity
for technological experiment
Current standard interdisciplinary treatment for stroke
rehabilitation is labor intensive, usually relying on oneon-one, manual interactions with therapists. Studies that
use enhanced training are no exception. The treatment
protocols rely on daily interaction over periods of weeks.
For stroke, because the therapy that promotes the best
Robot-aided sensorimotor training Volpe et al. 747
recovery is unknown, most therapists use a combination
of traditional techniques. Patient evaluation is usually
done subjectively, making it dif®cult to monitor treatment effects. This situation presents an opportunity to
create new technological solutions to the problems of
neuro-recovery. Devices that provide safe, quanti®able,
and reproducible physical activity would clearly assist
health care delivery experts.
Whether experiments with these devices produce added
value are challenges that loom large. The primary
challenge is whether the robotic training has ef®cacy ±
does it work? When compared with control individuals,
does enhanced sensorimotor training with robotic
devices produce not only decreased impairment but also
decreased disability, and are the motor improvements
long lasting. Second, and for the present beyond the
scope of the present review, if robotic training is
effective then are cost ef®ciencies obtained as a result?
Recent data gathered by several groups concentrate on
the use of upper limb robotics in patients with stroke.
Those studies have proposed initial design standards,
and have demonstrated preliminary results in over 100
patients who have had robotic sensorimotor therapy
added to standard rehabilitation programs. We consider
those data as an example of the potential of technologybased methods, and intentionally leave to another
discussion the complementary approaches of other
devices that are meant to enhance recovery of gait.
Such approaches include body weight supported treadmill trainers [27,28,29 .,30,31 .,33 .], functional neuromuscular stimulation for gait training [34 .,35 .], and other
functional electrical stimulation strategies for upper limb
recovery [36,37].
Robot training: tools for therapists to
increase productivity and quality of care
The idea we are attempting to test is not whether robots
or robotic devices can serve the brain injured patient, the
`tin man' metaphor. Rather, we are testing whether we
can improve motor outcome by equipping therapists
with robotic devices to enhance the sensorimotor aspect
of rehabilitation. Clinical and scienti®c rationales exist
for added sensorimotor training during the recovery
period after stroke.
Before examining the outcome results from several
groups, it is worth discussing the practical advantages
and disadvantages of robotic training. On the positive
side, the robot will deliver a quanti®able input and
measure the patient's output objectively. This
unparalleled objective measurement of movement
will contribute to other outcomes research, for
example combined robotic and pharmacological intervention. A standard day in a rehabilitation hospital
requires at least 3 h of instruction, leaving several
time periods available for additional therapy. Under
optimal conditions and assuming the effect on
outcome is positive, a single therapist could manage
multiple patients, each of whom is interacting with
their device. Patient acceptance of the robotic
devices as training machines ®ts into their conception
of the standard rehabilitation gym that is already
®lled with wheel and pulley devices, and patients'
general interest in the multimedia aspect of some of
the devices is also compelling. The effectiveness of
the robotic training program with regard to impairment and disability reduction is yet to be fully
determined. The measure of the cost is a more
dif®cult question, and should be postponed pending
the determination of effectiveness.
It is important to note that there are at least two different
approaches to the application of robotic technology.
Brie¯y, industrial robots typically emphasize controlling
robot position, so that applications that do not require
controlled contact (e.g. automobile spray-painting) comprise its most successful and safe application. Reliable
control of the forces exerted on a manipulated object is
more challenging, although it has been shown to be
practicable. We have taken an alternative approach of
controlling the dynamic relation between motion and the
forces of interaction between the robot and the object it
manipulates [38±41]. (Material published before the
annual period of review provides the initial articulations
of the impedance control concept [38,39], a patent
statement [40], and key engineering references that
motivated the low-impedance, high-compliance approach [41].) This strategy with its advantage of rapidly
and smoothly responding to forces exerted on it, so that
it is compliant and easy to move, emphasizes a
`biologically friendly' quality [39,42]. (In an important
historical article, Flash and Hogan [42] argued that the
`minimum jerk' hypothesis captured signi®cant information about movement generation.) Along these lines,
Krebs et al. [43] found that initial movements as recovery
proceeded conformed to the minimum jerk hypothesis,
and indicated that the recovering patients blended
submovements in performing the motor task. These
`back-driveable' machines also tolerate rapid or uncontrolled movement, as occurs in tremor or myoclonus,
with a safety factor that exceeds that of the industrial
designed position controlled machines [43]. Whether
one strategy is more clinically effective than another is
an empirical issue. It may be that all approaches are
successful in particular groups of patients.
Experimental results on robotic
enhanced rehabilitation
The Burke±Massachusetts Institute of Technology
group subjected the upper limb robotic device to a test
748 Trauma and rehabilitation
of clinical effectiveness for the ®rst time some 6 years
ago. The experimental design has not changed, although
the designs to explore aspects of motor recovery that
have characteristics of motor learning have evolved, and
new efforts have focused on the development of the
visual display in order to motivate interest in exercise.
Figure 1 depicts a typical patient setup in front of the
robotic device for treatment. Consecutive patients were
randomly assigned to a robot treatment or control group.
Inclusion criteria were that the patients had been
admitted to the rehabilitation hospital within 3 weeks
of their ®rst stroke, had sustained some upper limb
weakness, and were able to follow a few simple
instructions. A `measuring' therapist who was unaware
of the patient's group assignment made all clinical
measurements of impairment and disability. All patients,
robot treated and control, underwent a standard interdisciplinary rehabilitation program, the robot treatment
or control was added to the program.
Robot training took place in a standard therapy suite
(supervised by a research therapist), lasted 45 min
(without setup, which took approximately 10 min), and
required that a patient performed 1024 ¯exion and
extension movements of the arm with gravity eliminated
in eight directions represented by the points of a
compass. The training program occurred 5 days per
week for 4 weeks. Control patients had less time on the
robot (1 or 2 h/week), the motors were never turned on,
and the patients moved the affected limb with the
unaffected limb.
Table 1 indicates the interval change from rehabilitation
admission to discharge in 96 patients. In order to
address the question posed in the central argument of
this paper, these data are a compilation of the past
published results with the current addition of 20 new
patients [44,45 . .,46,47,48 . .,49 . .]. Details of the new
adaptive training protocols and the various temporal
sequences will be elaborated elsewhere. The robottrained group demonstrated increased Fugl±Meyer for
shoulder and elbow (maximum 42) scores, but the
difference was not signi®cant. The expanded measure
of upper limb motor impairment (Motor Power,
standard muscle testing, maximum 20) and the expanded mixed measure of upper limb impairment and
disability (the Motor Status Score for shoulder and
elbow, maximum 40) were signi®cantly improved in the
robot-trained group as compared with the control group.
In data reported elsewhere [44,45 . .,46], the motor
improvement was con®ned to the exercised proximal
limb: movements around the shoulder and elbow.
There was no advantage conferred by robotic training
on sensorimotor activity of the wrist and ®ngers. Motor
performance of lower limb activity, especially gait, was
comparable between groups.
Figure 1. Robot training with the MIT-Manus at the Burke Medical
Research Institute
A patient seated in front of the Massachusetts Institute of Technology±
Manus device with her shoulders strapped to the chair and moving the
manipulandum. The patient's hand is strapped to a wrist carrier attached
to the manipulandum. The video screen is above the training table.
Table 1. Change from rehabilitation admission to discharge in robottrained and control stroke patients
Group
Robot-trained
(n = 56)
Control (n = 40)
Significance
FM S/E
(maximum 42)
MSS S/E
(maximum 40)
MP
(maximum 20)
6.6+1.0
4.9+0.8
NS
8.6+0.9
3.4+0.5
P50.001
4.1+0.4
2.2+0.3
P50.005
These 96 patients with stroke had rehabilitation treatment shortly after
the event, and the robot-trained group demonstrated significant
improvement as compared with the control patients. The impairment
measurements depict interval change (mean+standard error). Timing of
stroke to rehabilitation (around 2.5 weeks), duration of rehabilitation
experience (around 3.5 weeks) and all admission impairment measures
were comparable between groups. FM S/E, Fugl±Meyer Score for
shoulder and elbow; MP, Muscle Power; MMS S/E, Motor Status Scale
for shoulder and elbow.
Figure 2 shows the impairment results on the Motor
Power Scale for the 96 patients, and follow-up evaluation
about 3 years after stroke for 31 patients. The robottrained group demonstrated sustained impairment gains
in the upper limb compared with the control individuals,
although the difference is not signi®cant; this indicates
that improvement continues, albeit to a modest degree,
many months after stroke.
For those 96 patients the interval change in the
Functional Independence Measure (a reliable disability
score) was comparable across groups. The Functional
Independence Measure has reliability and validity, but
many of the activities measured in the self-care
subsection that depend on upper limb motor control
Robot-aided sensorimotor training Volpe et al. 749
Figure 2. Change in motor power after robot training or control in
patients recovering from stroke
Motor 20
power 17.5
15
12.5
10
7.5
5
2.5
0
Admission
(n = 96)
Discharge
(n = 96)
Follow up
(n = 31)
Control
Robot-trained
Mean+standard error Motor Power scores (maximum score 20) of 96
patients on admission before rehabilitation, at discharge after rehabilitation and robotic training or control, and at follow-up evaluation
approximately 3 years after stroke. Robot-trained patients maintain the
motor improvements. *P50.05, versus control.
can be performed with one limb. Because these scores
were not obtained with the restriction of using only the
paralyzed or unaffected limb, a de®nitive conclusion
regarding the links among decreased upper limb motor
impairment, increased motor function and decreased
disability is not currently possible.
The Palo Alto Veteran's Affairs±Stanford group has
pioneered the use of robotic training for patients who
had stroke for many months or years [50,51,52 . .,53 .,54].
Their published results focus on treating over 24
patients 6 months to 2 years after stroke with a robotic
device called `MIME' (mirror image motion enabler; a
PUMA 260 machine). Patients were randomly assigned
to robot treatment or control. All patients were in a
treatment program for the same amount of time. Robottreated patients had their upper extremity manipulated
by the robot as well as the therapist, whereas control
patients had only the therapist manipulate their affected
upper limb. Therapists determined the stage and the
application of the individually tailored robot treatment
protocol.
The results demonstrate that the robot-treated patients
had signi®cantly greater interval change in the Fugl±
Meyer scale score for shoulder and elbow activity, but
not for wrist and hand activity [52 . .,53 .,54]. The robottrained patients also demonstrated signi®cantly improved percentage change in mean strength of shoulder
and elbow movements as compared with control
patients. Consistent with the Burke±Massachusetts
Institute of Technology results in patients treated
within weeks of stroke, the Palo Alto Veteran's
Affairs±Stanford experiments demonstrated that recovery may be enhanced during different time periods,
because it almost certainly continues in small increments months to years after the acute stroke. Additional
recent work from that group explored the added effect
of visual context on stroke recovery. They designed a
split wheel robotic device that requires different levels
of force on various parts of the wheel to train and test
driving ability [54 . .]. Other groups have also explored
the use of the visual display to increase interest and
motivation [47,55].
The Rehabilitation Institute of Chicago±University of
California at Irvine group reported the initial training
results of adding a therapy with a robotic controlled
reaching device (Assistive Rehabilitation and Measurement Guide) to a standard program [56,57 . .]. (Reinkensmeyer et al. [56] provided engineering information
for the device, which can guide upper limb movement against gravity). They trained over 12 patients
with chronic stroke (some over 5 years after stroke)
on the reaching paradigm. Initial results demonstrated
that trained patients had improved kinematics of reach
and velocity, and better control of tone; patients
produced smoother movements [57 . .]. If smoothness
or the quality of movement acquired in recovery
matters to outcome, and smoothness has been shown
to be a de®ning characteristic of normal, coordinated
movement (42 for example), then these detailed
measurements obtained only by technological instrumentation will add another clinically important dimension. A randomized study [58,59 . .] demonstrated that
control patients treated with an equal number of
movements directed by a therapist improved to a
level comparable to the level achieved by those
trained using the Assistive Rehabilitation and Measurement Guide device. Furthermore, using a disability measure of bimanual functional activity, those
investigators found trends that favored the robottreated group [60].
Conclusion
A variety of robotic approaches also appear to in¯uence
favorably the motor outcome in patients with chronic
injury. In combination, these data demonstrate that
robotic sensorimotor training added to an acute rehabilitation program improved motor outcome as compared
with control patients. These studies were prospective
and randomized, with masked therapists acting to assess
outcome [44±49]. The improved outcome appears
concentrated in the exercised limb. In the follow up
studies to date, the advantage conferred by robotic
training continues for at least 3 years. These promising
results prompt further questions.
750 Trauma and rehabilitation
In addition to the modest gains in impairment
reduction, some groups are identifying gains in disability reduction. Although some of the current robotic
devices could be programmed to teach a patient to
compensate for a task-related de®cit, the current
protocols were not designed to compensate for a
speci®c task. The broader goal is to help therapists
treat impairment more effectively. Whether these
devices will replace therapists or exist as a potent new
tool for therapists should settle de®nitively on the latter
purpose. For the robot to be an effective tool, gains in
impairment should translate into disability gains. In
order to make a paralyzed arm more functional,
improved wrist, hand, and ®nger function must follow.
Robotic devices are currently in the test phase that will
train the wrist and ®ngers, and continue to develop
shoulder elevation [61±64]. The impairment in the
distal upper limb may dictate whether robotic training
programs can in¯uence disability. These tests will be
completed soon, and results will become available in
the near future.
Experiments should determine whether the critical
ingredient that underlies motor improvement is the
intensity of the movement experience. An approach to
this question requires that the treatment and control
group spend equivalent time, respectively, moving or
being moved by a device or by a therapist. The
Rehabilitation Institute of Chicago±University of California at Irvine group has shown that comparable training
by therapists (and speci®cally the same number of
movements [58 .]) leads to comparable clinical outcomes
between robot-trained and control patients. Under the
circumstances of the amount of movement that is
possible on a robotic device, it may be dif®cult for the
typical one-on-one therapist±patient interaction to compete. For example with the Burke±Massachusetts
Institute of Technology device (the Manus), a subject
typically had over 20 000 extra ¯exion±extension upper
limb movements. Nevertheless, there are wheel and
pulley devices that could be employed to test this
question.
There should be a collaborative effort to gather the data
with alacrity. The groups should agree on subjective
clinical scales so that they are measuring the same
movement variables, and the effectiveness of devices
could be compared. New measurement scales must have
bona ®de reliability and correlate with the long established, reliable scales. There should be an attempt to
generate objective measures and test the correlation of
these objective measures with the standard clinical
scales. Finally, there should be a concerted effort to
understand the relationship between structure and
function. Clearly, some patients with stroke undergo
better recovery than others.
There are a variety of robotic techniques and nonrobotic
techniques that have demonstrated effectiveness in
changing the level of impairment, and some have
in¯uenced the level of disability. For the ®rst time, to
our knowledge, these data demonstrate that more
sensorimotor training leads to better motor outcome
compared with control. One of the possible reinterpretations of the question in the title of the present review
might be, `Can we afford to continue emphasizing
disability reduction at the expense of a balanced
approach to impairment, as well as disability reduction?'
A ®nal determination regarding whether the use of
robotic devices in stroke rehabilitation is realistic needs
to be postponed, because more work is required.
However, arguments have been advanced that the
proper question might be whether neglecting to arm
the therapist with new tools, among them robotic
devices, can continue to be a realistic option.
Acknowledgements
The authors acknowledge support from the Burke Medical Research
Institute, the USPHS (NIH, HD 37397 and 36827), and the Langeloth
Foundation.
References and recommended reading
Papers of particular interest, published within the annual period of review, have
been highlighted as:
.
of special interest
..
of outstanding interest
1
American heart disease: 2001 heart and stroke statistical update. 2001.
p. 16.
2
Broderick J, Brott T, Kothari R, et al. The Greater Cincinnati/Northern
Kentucky Stroke Study: preliminary first-ever and total incidence rates of
stroke among blacks. Stroke 1998; 29:415±421.
3
Taub E, Miller NE, Novack TA, et al. Technique to improve chronic motor
deficit after stroke. Arch Phys Med Rehab 1993; 74:347±354.
4
Dromerick AW, Edwards DF, Hahn M. Does the application of constraintinduced movement therapy during acute rehabilitation reduce arm impairment
after ischemic stroke? Stroke 2000; 31:2984±2988.
5
Miltner WH, Bauder H, Sommer M, et al. Effects of constraint induced
movement therapy on patients with chronic motor deficits after stroke: a
replication. Stroke 1999; 30:586±592.
6
Liepert J, Bauder H, Wolfgang HR, et al. Treatment-induced cortical
reorganization after stroke in humans. Stroke 2000; 31:1210±1216.
7
Kwakkel G, Wagenaar RC, Twisk JWR, et al. Intensity of leg and arm training
after primary middle cerebral artery stroke: a randomized trial. Lancet 1999;
354:191±196.
8
Kwakkel G, Wagenaar RC, Koelman TW, et al. Effects of intensity of
rehabilitation after stroke: a research synthesis. Stroke 1997; 28:1550±1556.
9
Feys HM, De Weerdt WJ, Selz BE, et al. Effect of a therapeutic intervention
for the hemiplegic upper limb in the acute phase after stroke: a single-blind,
randomized, controlled multicenter trial. Stroke1998; 29:785±792.
10 Sunderland A, Fletcher D, Bradley L, et al. Enhanced physical therapy for
arm function after stroke: a one year follow up study. J Neurol Neurosurg
Psychiatry 1994; 57:856±858.
11 Sunderland A, Tinson DJ, Bradley L, et al. Enhanced physical therapy
improves recovery of arm function after stroke. A randomized clinical trial. J
Neurol Neurosurg Psychiatry 1992; 55:530±535.
12 Cao Y, D'Olhaberriague L, Vikingstad EM, et al. Pilot study of functional MRI
to assess cerebral activation of motor function after poststroke hemiparesis.
Stroke 1998; 29:112±122.
13 Cramer SC, Bastings EP. Mapping clinically relevant plasticity after stroke.
Neuropharmacology 2000; 39:842±851.
Robot-aided sensorimotor training Volpe et al. 751
14 Cramer SC, Finklestein SP, Schaechter JD, et al. Distinct regions of motor
cortex control ipsilateral and contralateral finger movements. J Neurophysiol
1999; 81:383±387.
15 Cramer SC, Moore CI, Finklestein SP, Rosen BR. A pilot study of
somatotopic mapping after cortical infarct. Stroke 2000; 31:668±671.
16 Miyai I, Suzuki T, Mikami A, et al. Patients with capsular infarct and Wallerian
degeneration demonstrate persistent regional premotor cortical activation on
functional MRI. Stroke 2001; (in press).
17 Nelles G, Jentzen W, Jueptner M, et al. Arm training induced brain plasticity in
.
stroke studied with serial positron emission tomography. NeuroImage 2001;
13:1146±1154.
Ten patients with stroke were randomly assigned to treatment with intensive upper
limb activity or standard therapy. The outcome trend favored the extensive
treatment group. Post-treatment online passive movement positron emission
tomography study demonstrated increased activation in bilateral inferior parietal
cortex, premotor cortex, and contralateral sensorimotor cortex for the enhanced
treatment group as compared with control patients and normal individuals.
33 Kosak MC, Reding MJ. Comparison of partial weight supported treadmill gait
.
training versus aggressive bracing assisted walking post stroke. Neurorehabil
Neural Repair 2000; 14:13±19.
In patients with the most severe impairment after stroke there was a significant
dose-related (more than 12 sessions) improvement, as measured by gait
endurance and speed for the gait-trained group.
34 Daly JJ, Ruff RL. Electrically induced recovery of gait components for older
.
patients with chronic stroke. Am J Phys Med Rehabil 2000; 79:349±360.
Two patients 1 and 4 years following stroke were treated with standard therapy
followed by implantation of electrodes (for delivery of functional neuromuscular
stimulation). Gains were recorded following standard treatment, and further gains
were recorded following functional neuromuscular stimulation. This again provides
evidence for salutatory patterns of recovery: plateaus followed by change.
35 Daly JJ, Ruff RL, Haycook K, et al. Feasibility of gait training for acute stroke
.
patients using FNS with implanted electrodes. J Neurol Sci 2000; 179:103±
107.
Implanted electrodes plus a stimulator and software that provided individualized
stimulation patterns, with flexible stimulus parameters and activation timings of
multiple muscles, were studied. Results demonstrated improved gait and
functional outcome in five patients with stroke.
18 Marshall RS, Perera GM, Lazar RM, et al. Evolution of cortical activation during
.
recovery from corticospinal tract infarction. Stroke 2000; 31:656±661.
Eight patients with stroke were studied with serial functional magnetic resonance
imaging (sequential thumb finger opposition task). Compared with normal control
individuals and their first functional magnetic resonance imaging, stroke patients
on second imaging demonstrated altered ratios of hemispheric activation. Mirror
movements present in the first functional magnetic resonance imaging were said to
be `rare' in the second.
37 Powell J, Pandyan AD, Granat M, et al. Electrical stimulation of wrist
extensors in poststroke hemiplegia. Stroke 1999; 30:1384±1389.
19 Jones T, Shallet T. Use dependent growth of pyramidal neurons after neocortical damage. J Neurosci 1994; 14:2140±2152.
38 Hogan N. Impedance control of industrial robots. In: Robotics and computerintegrated manufacturing, vol. 1, no. 1. 1984. pp. 97±113.
20 Kozlowski D, James DC, Shallert T. Use dependent exaggeration of neuronal
injury after unilateral sensorimotor cortex lesions. J Neurosci 1996; 16:4776±
4786.
39 Hogan N. Adaptive control of mechanical impedance by coactivation of
antagonist muscles. IEEE Trans Automatic Control 1984; AC-29:681±690.
21 Kozlowski DA, Schallert T. Relationship between dendritic pruning and
behavioral recovery following sensorimotor cortex lesions. Behav Brain Res
1998; 97:89±98.
22 Nudo RJ, Wise BM, SiFuentes F, Milliken GW. Neural substrates for the
effects of rehabilitative training on motor recovery after ischemic infarct.
Science 1996; 272:1791±1794.
36 Chae J, Bethoux F, Bohine T, et al. Neuromuscular stimulation for upper
extremity motor and functional recovery in acute hemiplegia. Stroke 1998;
29:975±979.
40 Hogan N, Krebs HI, Sharon A, Charnnarong J. Interactive robot therapist. US
Patent #5,466,213,MIT; November 14, 1995.
41 Hogan N. Impedance control: an approach to manipulation. Part 1, 2, 3
ASME. J Dyn Syst Measure Control 1985; 107:1±24.
42 Flash T, Hogan N. The coordination of arm movements: an experimentally
confirmed mathematical model. J Neurosci 1985; 5:1688±1703.
23 Nudo RJ, Milliken GW, Jenkins WM, Merzenich MM. Use-dependent
alterations of movement representations in primary motor cortex of adult
squirrel monkeys. J Neurosci 1996; 16:785±807.
43 Krebs HI, Aisen ML, Volpe BT, Hogan N. Quantization of continuous arm
motion in humans with brain injury. Proc Natl Acad Sci USA 1999; 96:4645±
4649.
24 Plautz EJ, Milliken GW, Nudo RJ. Effects of repetitive motor training on
movement representations in adult squirrel monkeys: role of use versus
learning. Neurobiol Learn Mem 2000; 74:27±55.
44 Aisen ML, Krebs HI, Hogan N, et al. The effect of robot assisted therapy and
rehabilitative training on motor recovery following stroke. Arch Neurol 1997;
54:443±446.
25 Schallert T, Kozlowski DA, Humm JL, Cocke RR. Use-dependent structural
events in recovery of function. Adv Neurol 1997; 73:229±238.
45 Volpe BT, Krebs HI, Hogan N, et al. A novel approach to stroke rehabilitation:
. . robot aided sensorimotor stimulation. Neurology 2000; 54:1938±1944.
Fifty-six patients (30 in treatment group, 26 control patients) were treated within 3
weeks after stroke. Control patients, measuring therapist, and design were
comparable to those of the first study [44]. A significant advantage in motor power
improvement as compared with control patients was identified, but there was no
effect on disability. Clinical characteristics and motor scores on admission,
location and size of stroke were comparable between groups. This work provides
preliminary evidence that, among patients with larger lesions, those who undergo
this form of treatment improve more than do control patients.
26 Xerri C, Merzenich MM, Peterson BE, Jenkins W. Plasticity of primary
somatosensory cortex paralleling sensorimotor skill recovery from stroke in
adult monkeys. J Neurophysiol 1998; 79:2119±2148.
27 Barbeau H, McCrea DA, O'Donovan MJ, et al. Tapping into spinal circuits to
restore motor function. Brain Res 1999; 20:27±51.
28 Barbeau H, Ladouceur M, Norman KE, et al. Walking after spinal cord injury:
evaluation, treatment and functional recovery. Arch Phys Med Rehab 1999;
80:225±235.
29 Edgerton VR, deLeon RD, Harkema SJ, et al. Retraining the injured spinal cord.
.
J Physiol 2001; 533:15±22.
Edgerton et al. provide a review of an exciting rationale for device intervention and
novel potential pharmacologic intervention. A companion paper [26,27] provides a
review of critical physiology.
30 Hesse S, Uhlenbrock D, Werner C, Bardelben A. A mechanized gait trainer
for restoring gait in nonambulatory subjects. Arch Phys Med Rehab 2000;
81:1158±1161.
31 Hesse S, Uhlenbrock D. A mechanized gait trainer for restoration of gait. J
.
Rehabil Res Dev 2000; 37:701±708.
This gait trainer has an updated design compared with a device that was
previously used successfully to restore gait in hemiplegic patients. Companion
paper [30] reports individual patients trained in an advanced system that `does not
overstrain' therapists. The later work also reports electromyography comparisons
with normal persons, suggesting improved gait kinematics for the patient while on
the trainer.
32 Colombo G, Joerg M, Schreier R, Dietz V. Treadmill training of paraplegic
.
patients using a robotic orthosis. J Rehabil Res Dev 2000; 37:693±700.
In this report, a treadmill trainer is described with driven gait orthosis that moves
the subject's legs.
46 Volpe BT, Krebs HI, Hogan N, et al. Robot training enhanced motor outcome
in patients with stroke maintained in three year follow-up. Neurology 1999;
53:1874±1876.
47 Krebs HI, Hogan N, Aisen ML, Volpe BT. Robot aided neuro-rehabilitation.
IEEE Trans Rehabil Eng 1998; 6:75±87.
48 Krebs HI, Volpe BT, Aisen ML, Hogan N. Increasing productivity and quality of
. . care: robot-aided neurorehabilitation. J Rehabil Res Dev 2000; 37:639±652.
Lesion location affected accuracy in a reaching task. Patient with cortical plus
striatal lesions had speedy but impaired accuracy on reaching movement as
compared with a patient with a striatal lesion alone who had slow but accurate
reaching movements. This paper emphasizes the potential use of robots as tools
that therapists could use to deliver individualized training to more than one patient.
The therapist's role may evolve away from labor-intensive manual treatment to a
supervisory and decision-making role.
49 Krebs HI, Volpe BT, Palazzolo J, et al. Robot aided neuro-rehabilitation in stroke:
. . interim results on follow-up of 76 patients and on movement indices. In:
Integration of assistive technology in the information age. Mokhtari M (editor).
Amsterdam, The Netherlands: IOS Press; 2001. pp. 45±59.
The robot-treated group had an outcome advantage with respect to motor
improvement. Additional evidence is provided that early movements after initial
complete paralysis appear to be blended (this is consistent with earlier work by
this group in patients with stroke [43].
752 Trauma and rehabilitation
50 Lum PS, Reinkensmeyer DJ, Lehman SH. Robotic assist devices for
bimanual physical therapy: preliminary experiments. IEEE Trans Rehab Eng
1993; 1:185±191.
55 Rose FD, Attree EA, Brooks BM, et al. Training in virtual environments:
transfer to real world tasks and equivalence to real world training.
Ergonomics 2000; 43:494±511.
51 Lum PS, Burgar CG, Kenney DE, Van der Loos HF. Quantification of force
abnormalities during passive and active-assisted upper-limb reaching movements in post-stroke hemiparesis. IEEE Trans Biomed Eng 1999; 46:652±
662.
56 Reinkensmeyer DJ, Dewald JPA, Rymer WZ. Robotic devices for physical
rehabilitation of stroke patients; fundamental requirements, target therapeutic
techniques and preliminary designs. Technol Disability 1996; 5:205±215.
52 Burgar CG, Lum PS, Shor PC, Machiel Van der Loos HF. Development of
..
robots for rehabilitation therapy: the Palo Alto VA/Stanford experience. J Rehabil
Res Dev 2000; 37:663±673.
This paper describes findings with a redesigned MIME using an industrial-type
robot (PUMA 560). Twenty-one patients (11 robot trained and 10 control patients),
who were at least 6 months poststroke (average 2 years), were enrolled. Patients
were all treated in the same treatment space and all were aided by either therapist
or robot. In is unclear how decisions were made regarding whether patients
received robot-aided reach therapy or therapist-aided reach therapy, or the
quantity of one or the other treatment. The authors claim to have delivered
comparable treatment intensity and duration to the two groups. Results indicate
that the treated group improved on standard impairment scores, but strength
(maximum voluntary isometric contraction) areas not treated (i.e. hand and wrist)
did not improve. Improved performance in the robot treated group was
accompanied by better activation patterns, measured as the difference between
pre- and post-treatment electromyography activations.
57 Reinkensmeyer DJ, Kahn LE, Averbuch M, et al. Understanding and treating
..
arm movement impairment after chronic brain injury: progress with the ARM
guide. J Rehabil Res Dev 2000; 37:653±662.
Three patients were included, who were 0.5, 2, and 5 years poststroke. The
severity of stroke was variable, but all patients demonstrated improvement in
reach, velocity, tone and off-axis force. Two-thirds improved reach to targets.
53 Shor PC, Lum PS, Burgar CG, et al. The effect of robot aided therapy on upper
.
extremity joint passive range of motion and pain. In: Integration of assistive
technology in the information age. Mokhtari M (editor). Amsterdam, The
Netherlands: IOS Press; 2001. pp. 79±83.
Twenty-seven patients (13 robot trained and 14 control), who had suffered a
stroke more than 6 month previously, were enrolled. The control patients were
treated with NDT; and the treatment regimen involved progression through four
modes; the time spent in each mode was not indicated. There were comparable
training times in both groups (24 1-h sessions). Early time points demonstrated an
outcome advantage for the treatment group, but control patients caught up by 8
months (change on the order of 2-4 points on the Fugl±Meyer scale). Results
demonstrated that low compliance systems (such as MIME) are safe, and there
was a trend toward less pain in the affected limb in the treatment group (both
results similar to the high compliance MIT-Manus device, 44).
54 Johnson MJ, Van der Loos HFM, Burgar CG, et al. Designing a robotic stroke
..
therapy device to motivate use of the impaired limb. In: Integration of assistive
technology in the information age. Mokhtari M (editor). Amsterdam, The
Netherlands: IOS Press; 2001. pp. 123±132.
Eight patients, long after they had sustained a stroke (5 years), but no control
patients were included; rather, patients were compared with normal individuals.
Nifty use of context and cues was employed (embedded corrective force cuing).
Patients faced a steering wheel, and viewed a driving scene and road display.
(Video games have arrived in neuro-rehabilitation.) The measure of limb power was
torque on a steering wheel; weak limbs move better with rather than against
gravity. No measure of generalization to other tasks was taken, or was a direct
measure of impairment or power.
58 Kahn L, Averbuch M, Rymer WZ, Reinkensmeyer DJ. Comparison of robot
.
assisted reaching to free reaching in promoting recovery from chronic stroke. In:
Integration of assistive technology in the information age. Mokhtari M (editor).
Amsterdam, The Netherlands: IOS Press; 2001. pp. 39±44.
Ten patients with stroke were randomized to robot treatment and comparable
training by a therapist. Results demonstrated that both groups improved with
regard to disability (Chedoke and Rancho Los Amigo); there were similar
improvement across groups for kinematic measures.
59 Kahn LE, Zygman ML, Rymer WZ, Reinkensmeyer DJ. Effect of robot-assisted
..
exercise on functional reaching in chronic hemiparesis. In: Proceedings of the
23rd Annual International Conference of the IEEE-EMBS; 25±28 October
2001; Istanbul, Turkey.
Fourteen patients with stroke were randomized to robot treatment and comparable
training by a therapist. Results demonstrated significant improvement in movement
smoothness for the robot-trained group.
60 Gowland C, Stratford P, Ward M. Measuring physical impairment and
disability with the Chedoke-McMaster Stroke Assessment. Stroke 1993;
24:58±93.
61 Krebs HI, Buerger SP, Jugenheimer KA, et al. 3-D extension for MITMANUS: a robot-aided neuro-rehabilitation workstation. ASME 2000 IDETC/
CIE, DETC2000/MECH-14151, September 2000.
62 Buerger SP, Krebs HI, Hogan N. Characterization and control of a screwdriven robot for neurorehabilitation. IEEE CCA/ISIC 2001, CCA-388,
September 2001.
63 Jugenheimer KA, Hogan N, Krebs HI. A robot for hand rehabilitation: a
continuation of the MIT-MANUS neuro-rehabilitation workstation. ASME
2001 IDETC/CIE, DETC2001/DAC-21085, September 2001.
64 Williams DJ, Krebs HI, Hogan N. A robot for wrist rehabilitation. In:
Proceedings of the 23rd Annual International Conference of the IEEE-EMBS;
25±28 October 2001; Istanbul, Turkey.