636 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 23, NO.

4, JULY 2015

Effects of Innovative WALKBOT Robotic-Assisted

Locomotor Training on Balance and Gait Recovery
in Hemiparetic Stroke: A Prospective, Randomized,
Experimenter Blinded Case Control Study With
a Four-Week Follow-Up
Soo-Yeon Kim, Li Yang, In Jae Park, Eun Joo Kim, Min Su JoshuaPark, Sung Hyun You,
Yun-Hee Kim, Hyun-Yoon Ko, and Yong-Il Shin

Abstract—The present clinical investigation was to ascertain effects of the WALKBOT-assisted locomotor training on balance,
whether the effects of WALKBOT-assisted locomotor training gait and motor recovery when compared to the conventional
(WLT) on balance, gait, and motor recovery were superior or sim- locomotor training alone in patients with hemiparetic stroke.
ilar to the conventional locomotor training (CLT) in patients with
Index Terms—Balance, gait, neurorehabilitation, robotic-as-
hemiparetic stroke. Thirty individuals with hemiparetic stroke
sisted locomotor training, stroke.
were randomly assigned to either WLT or CLT. WLT emphasized
on a progressive, conventional locomotor retraining practice (40
min) combined with the WALKBOT-assisted, haptic guidance
and random variable locomotor training (40 min) whereas CLT I. INTRODUCTION
involved conventional physical therapy alone (80 min). Both inter-

S
vention dosages were standardized and provided for 80 min, five TROKE is a common, leading cause of balance and loco-
days/week for four weeks. Clinical outcomes included function
ambulation category (FAC), Berg balance scale (BBS), Korean motor disorders that may lead to chronic physical disabil-
modified Barthel index (K-MBI), modified Ashworth scale (MAS), ities; patients with stroke are burdened with high medical costs
and EuroQol-5 dimension (EQ-5D) before and after the four-week [1], [2]. They are vulnerable to balance and locomotor dysfunc-
program as well as at follow-up four weeks after the intervention. tions due to the hemiparetic lower extremity after a sudden loss
Two-way repeated measure ANOVA showed significant interac- of the brain functions. In addition, they are more inclined to use
tion effect (time group) for FAC , BBS ,
and K-MBI across the pre-training, post-training, nonaffected lower extremity while standing and walking. Thus,
and follow-up tests, indicating that WLT was more beneficial for they are vulnerable to arrhythmic, asymmetric weight bearing
balance, gait and daily activity function than CLT alone. However, and reduced gait cycle [3], [4]. Animal experimental studies
no significant difference in other variables was observed. This have shown that neuroplasticity leading to motor recovery in
is the first clinical trial that highlights the superior, augmented ischemic lesions occurs after the balance and gait training was
repeated 400–600 times [5], [6]. However, according to the re-
port of Lang et al., the typical numbers of repetitions for balance
Manuscript received September 14, 2014; revised December 03, 2014; ac-
and gait (steps) training accounted for 6.0 and 291.5 repetitions
cepted February 11, 2015. Date of publication April 02, 2015; date of current
version July 03, 2015. per each session, respectively, in a neurorehabilitation program
S.-Y. Kim, H.-Y. Ko, and Y.-I. Shin are with the Department of Reha- [7]. It can therefore be inferred that the effects of the conven-
bilitation Medicine, Pusan National University School of Medicine, Pusan
tional locomotor training would be enhanced if combined with
626-770, Korea, and also with the Research Institute for Convergence of
Biomedical Science and Technology, Pusan National University Yangsan the robot-assisted one providing an accurate sensorimotor feed-
Hospital, Pusan, Korea (e-mail: drkimsy@gmail.com; drkohy@gmail.com; back via haptic guidance and various built-in software programs
rmshin01@gmail.com).
in the early stage of stroke rehabilitation [8].
L. Yang is with the Department of Rehabilitation Medicine, Pusan
National University School of Medicine, Pusan 626-770, Korea (e-mail: Neurorehabilitation programs have been used to improve
lilyaihao@gmail.com). balance and locomotor functions; these include neurodevelop-
I. J. Park, E. J. Kim, and M. S. Park are with Research Institute for
mental technique, repetitive task training, biofeedback, body
Convergence of Biomedical Science and Technology, Pusan National Uni-
versity Yangsan Hospital, Pusan, Korea (e-mail: smartpark85@gmail.com; weight-supported treadmill training, robot-assisted training, and
10044861@naver.com; minsu.park.otr@gmail.com). high-intensity physical therapy. Still, however, controversial
S. H. You is with the Department of Physical Therapy Program, Yonsei Uni-
opinions exist regarding their effects in improving the balance
versity, Wonju, Korea (e-mail: joshuayou7@gmail.com).
Y.-H. Kim is with the Department of Physical and Rehabilitation Medicine, and locomotor functions [9], [10]. Of these, the task-specific
Stroke and Cerebrovascular Center, Samsung Medical Center, Sungkyunkwan locomotor training with a body weight-supported treadmill
University School of Medicine, Seoul, Korea (e-mail: yunkim@skku.edu).
has been commonly used in a clinical setting [11]–[14]. But
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. it is disadvantageous in that it is often labor-intensive, thus
Digital Object Identifier 10.1109/TNSRE.2015.2404936 posing challenging problems for clinicians who aim to raise

1534-4320 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
KIM et al.: EFFECTS OF INNOVATIVE WALKBOT ROBOTIC-ASSISTED LOCOMOTOR TRAINING ON BALANCE AND GAIT RECOVERY 637

the frequency of the repetition of the locomotor training for

neuroplasticity in the neurorehabilitation of patients with
stroke. To address these issues, we have recently developed a
robot-assisted locomotor training system, WALKBOT (P&S
Mechanics, Seoul, Korea), for the purpose of providing an
accurate haptic guidance via a proprioceptive, kinematic and
kinetic biofeedback, variable practice, high-intensity, repeti-
tive, task-specific and interactive exercises for patients with
paretic lower limbs [15]. It is useful to perform a quantitative
analysis of the kinematic and kinetic parameters and spastic
stiffness during the weight-supported treadmill walking and
then to provide a sensorimotor feedback for the patients [15]. Fig. 1. Study flowchart.
Recent clinical studies have shown that the robot-assisted
gait training system is highly effective in improving the lo-
comotor functions as compared with the conventional one; it
shows a new paradigm for the guidance practice [16]–[20].
According to a review of the literature, however, the guidance
practice and the variable one are predominantly involved in the
early and the intermeditate-to-late stage of the gait training,
respectively [15]. Thus, both practices are involved in the
long-term potentiation and neuroplasticity [21]. Recently, a
randomized, controlled study was conducted to assess the
effects of a 12-week robot-assisted gait training program using
the LOKOMAT (Hocoma AG, Volketswil, Switzerland) in
improving the gait velocity in patients with incomplete, chronic
spinal cord injury, thus demonstrating its minimal efficacy
[22]. Thus, it showed that the LOKOMAT was effective only
in a minimal manner presumably because it is only efficient
in adapting to the degree of the need for the gait training or
responding to that of the complexity [23]. It can therefore be
inferred that the robot-assisted gait training system might also
be effective in improving the gait velocity if combined with
conventional one that provides a contextual interference in the
gait training based on the variable practice or if installed with
the variable practice. To date, however, no studies have been
conducted to raise the gait velocity in patients with stroke.
Therefore, the specific aim of the present investigation was to
evaluate whether the effects of WALKBOT-assisted locomotor
Fig. 2. Side view of the Walkbot.
training (WLT) on balance, gait and motor recovery were su-
perior or similar to the conventional locomotor training (CLT)
in individuals with hemiparetic stroke. We hypothesized that tremor; 3) the patients with severe visual and cognitive im-
WLT would produce greater enhancements on balance, gait, pairments; 4) the patients with musculoskeletal diseases (e.g.,
and motor recovery than CLT alone in patients with hemiparetic arthritic pain); 5) the patients with cardiopulmonary diseases
stroke. (e.g., unstable angina or hypertension); 6) the obese patients
with a body weight of 135 kg); 7) The patients with a short
II. SUBJECTS AND METHODS height of 150 cm. Patients were assigned randomly to either
the WLT group or the CLT group.
All the patients submitted a written informed consent.
A. Subjects and Study Procedure
The current study was approved by the Institutional Review
In the current prospective, randomized, single-blind study, we Board (IRB) of Pusan National University Yangsan Hospital
enrolled a total of 30 patients (22 men, mean age (IRB approval number: # 03-2013-011). Moreover, the cur-
years) with stroke. Inclusion criteria for the current study rent study was registered with ClinicalTrials.gov (Identifier:
are as follows: 1) the patients with first stroke whose onset not NCT02053233).
exceeded one year; 2) the patients who reached a almost plateau
in recovery of the locomotor functions after a 30-day conven- B. Evaluation Scales
tional neurorehabilitation [24]. Exclusion criteria for the cur- In the current study, we evaluated the locomotor functions
rent study are as follows: 1) the patients with severe spasticity based on such scales as the Functional Ambulation Category
based on the modified Ashworth's scale 2; 2) the patients with (FAC) [25], Berg Balance Scale (BBS) [26], Korean version
638 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 23, NO. 4, JULY 2015

TABLE I
DEMOGRAPHIC AND BASELINE CLINICAL CHARACTERISTICS OF THE PATIENTS

Abbreviations: WLT, WALKBOT-assisted locomotor training; CLT, conventional locomotor training; FAC, Functional
Ambulation Category; BBS, Berg Balance Scale; K-MBI, Korean version of Modified Barthel Index; EQ-5D, EuroQol-5
dimension; and MAS, Modified Ashworth Scale.
Data are expressed as mean SD (SD: standard deviation).
-values at the independent t-test.

the spasticity using the MAS and the quality of health using the
EQ-5D. The validity and reliability and validity of each scale
have been well documented [27], [28].

C. Intervention
WLT emphasized on a progressive, conventional locomotor
retraining practice (40 min) combined with the WALKBOT-as-
sisted, haptic guidance and random variable locomotor training
(40 min) whereas CLT involved conventional physical therapy
alone (Fig. 1). Both intervention
dosages were standardized and provided for 40 min 2
times/day (total of 80 min), five days/week for four weeks.
Specifically, CLT was comprised of bed mobility exercises
Fig. 3. Enrollment and flow of subjects of the study. (rolling, bridging, quadruped) and stretching (5 min); training
of balance (i.e., maintaining, reactive, and anticipatory postural
control exercise) during sitting (5 min); training the transfer
of Modified Barthel Index (K-MBI), Modified Ashworth Scale from sit-to-stand and vice versa while maintaining static
(MAS), and EuroQol-5 dimension (EQ-5D). To do this, we mea- and dynamic balance and strengthening exercise for tibialis
sured scores of each scale at baseline, endpoint and follow-up. anterior, quadriceps and gluteus maximus and medius with
Thus, we evaluated the balance using the BBS, locomotor func- or without functional electrical stimulation (FES) (10 min);
tions using the FAC, daily activity function using the K-MBI, standing balance training with force, center of pressure, sway,
KIM et al.: EFFECTS OF INNOVATIVE WALKBOT ROBOTIC-ASSISTED LOCOMOTOR TRAINING ON BALANCE AND GAIT RECOVERY 639

TABLE II
FUNCTIONAL SCORES OF EACH EVALUATION SCALE AT BASELINE, FOUR WEEKS, AND EIGHT WEEKS

Abbreviations: WLT, WALKBOT-assisted locomotor training; CLT, conventional locomotor training; FAC, Functional Ambulation Category;
BBS, Berg Balance Scale; K-MBI, Korean version of Modified Barthel Index; EQ-5D, EuroQol-5 dimension; and MAS, Modified Ashworth Scale.
Data are expressed as mean SD (SD: standard deviation).
Statistical significance at repeated measures of analysis of variance (ANOVA).

symmetry of weight bearing, and position biofeedback using a Initially, continuous visual and proprioceptive feedback about
force platform with force sensors (10 min); treadmill locomotor sagittal kinematics and force trajectories of the hip, knee, and
training with the patient's body-weight partially supported by ankle joints, approximating symmetrical, rhythmic, reciprocal
a harness and progressing to overground gait training with or locomotor pattern were provided to stimulate corresponding
without assistive devices, orthotics, or FES (10 min) [29]. central pattern generators (CPG) network, which play impor-
For the WLT, in addition to the conventional physical therapy tant roles in creating the rhythm and shaping the pattern of the
described previously, the Walkbot-assisted gait training was motor neuron firings in the spinal cord [30]. However, as the
augmented for another 40 min. All individuals wore a suspen- locomotor skill becomes mostly rhythmic and automatic, more
sion vest and harness connected to a counterweight system to variable practice using different walking velocities and guid-
provide lumbopelvic stability and body weight support. The pa- ance forces were provided to maximize locomotor learning.
tient's hip, knee and ankle joint axes were consistently aligned Furthermore, the subject could increase or decrease walking
with the exsoskeletal system's actuators and elastic straps were velocity automatically during the gait training session as he or
used to secure the legs (Fig. 2). Unlike LOKOMAT-assisted she improved walking performance. Because the locomotion
training system, the WALKBOT assisted system has an in- is primarily mediated by spinal locomotor reflex or the CPGs
dependent ankle actuator to control excessive ankle plantar with a cortical or subcortical modulation, a subcortical motor
flexion and toe clearance. Depending on the neuromuscular learning paradigm was used where the subjects were instructed
skeletal conditions (e.g., pain, muscle weakness, spasticity, to kick a ball automatically in front of the treadmill frame rather
tolerance, fatigue, or endurance) of each patient, approximately than consciously attempting to make an accurate step [31].
40%–60% (adjustable range, 0%–100%) of the total body Blood pressure and heart rate were assessed and monitored
weight was initially supported at the first session, and then to maintain below 80% of age-appropriate level during the
gradually decreased in 5%–10% increments per session as training sessions and inter-training rest was provided as needed.
tolerated without substantial knee buckling or toe drag. Based
on each individual's height, stride length, and walking velocity D. Statistical Analysis
were concurrently adjusted at 1.0–1.6 m/cycle and at 1.00–1.20 All data was expressed as mean SD (SD: standard devia-
km/h during the initial session, respectively. The walking tion). We used two-way repeated measures analysis of variance
speed was increased by 0.1 km/h every 5 min as tolerated (ANOVA) to compare scores of each scale between baseline
to 2.40–2.60 km/h (maximally adjustable to 3.00 km/h), and and endpoint in each group. Then, we also used the independent
remained thereafter for subsequent visits. The guidance force t-test to compare differences in scores of each scale between the
or torque of the knee and hip actuators can be adjusted from two groups. Finally, we performed a post-hoc analysis to iden-
100% to 10% (with a 10% increment) for one leg at a time. tify pairwise differences. Statistical analysis was done using the
640 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 23, NO. 4, JULY 2015

TABLE III
KOREAN-MODIFIED BARTHEL INDEX SCORES OF EACH EVALUATION SUBSCALES AT BASELINE, FOUR WEEKS, AND EIGHT WEEKS

Abbreviations: WLT, WALKBOT-assisted locomotor training; CLT, conventional locomotor training.

Data are expressed as mean SD (SD: standard deviation).
Statistical significance at repeated measures of analysis of variance (ANOVA).

SPSS for windows version 12.0 (SPSS, Chicago, IL, USA). A , BBS , and K-MBI across the
-value of 0.05 was considered statistically significant. pretest, posttest, and follow-up tests. In the subscales analysis
of K-MBI, subscales of dressing and ambulation
III. RESULTS showed significant interaction effect and subscale of
transfer revealed moderately significant interaction
A. Demographic and Baseline Clinical Characteristics of the
effect (time group) across the pretest, posttest, and follow-up
Subjects
tests (Table III). There were no significant differences in mean
The 26 subjects have successfully completed the pretest, in- scores of the EQ-5D and MAS between the two groups
tervention, posttest, and follow up test, and attrition rate was .
13.33% at follow-up (Fig. 3). One subject dropped out because As shown in Fig. 4, a post-hoc analysis showed that WLT was
of rib fracture which was not related to the study and the other more beneficial for balance, gait, and daily activity function than
three subjects dropped out because of the decline in general CLT alone .
health condition. Baseline and clinical characteristics of the pa-
tients are represented in Table I. There were no significant dif- IV. DISCUSSION
ferences in the age, the male-to-female ratio, the time from the For the current study, we have hypothesized that WLT might
onset of stroke, the ratio of the left-to-right side and the ratio achieve a recovery of balance, gait, and motor functions to a
of ischemic-to-hemorrhagic stroke between the two groups greater extent as compared with CLT alone in subjects with
. hemiparetic stroke. Thus, we found that there were significant
differences in mean scores of the FAC, BBS, and K-MBI
B. Outcome Measures between the two groups. We therefore reached a conclusion
As shown in Table II, two-way repeated measure ANOVA that the robot-assisted gait training is more effective in im-
revealed significant interaction effect (time group) for FAC proving the balance, gait, and motor functions when combined
KIM et al.: EFFECTS OF INNOVATIVE WALKBOT ROBOTIC-ASSISTED LOCOMOTOR TRAINING ON BALANCE AND GAIT RECOVERY 641

Our results are consistent with the results of a randomized,

clinical study showing that a three-week course of robotic-as-
sisted gait training was more effective in improving the gait ve-
locity, endurability, muscle strength and muscle tone as com-
pared with the conventional one in 16 patients with hemiparetic
stroke [19]. Presumably, this might be because the WALKBOT-
assisted locomotor training system provides a haptic guidance
or a proprioceptive and somatosensory feedback by control-
ling an exoskeletal orthotic devices involved in the coordinated,
rhythmic, kinematic, and kinetic movement of the hip, knee,
and ankle. In association with this, it has been suggested that
afferent proprioceptive signals generated from the haptic guid-
ance stimulate the central pattern generator (CPG) in the net-
work of motor neurons in the spinal cord and thereby are in-
volved in the rhythmic, coordinated intersegemental locomotion
of the limb [32]. Moreover, the efficacy of the robot-assisted gait
training system is based on the body weight-supported treadmill
walking. To put this in another way, it is useful to provide a body
weight bearing for patients with early stoke without fear of falls
[33]. These benefits of robot-assisted gait training system were
made to improve the balance-related functions, such as BBS
score and transfer and ambulation subscales of K-MBI. In par-
ticular, it plays an important role that the patients overcome the
fear of falling and obtain the confidence about the walk in the
subacute period of stroke. Moreover, it is advantageous in that it
simulates the human locomotion. At a walking cadence of 100
steps/min and the robot-assisted gait training system provides a
20-min session of training, it allows patients to repeat the gait
training for the postural and locomotor control up to 2000 times.
As compared with the conventional gait training, the frequency
of the repetition of up to 2000 times is sufficient to provoke
plasticity of motor neurons and to achieve a recovery of loco-
motor functions [34]. In addition, the robot-assisted gait training
system is also useful to control the posture and locomotor func-
tions only in a limited scope and to provide a contextual inter-
ference by making minimal changes in them in response to the
intensity of the gait training and the coordination between the
two limbs.
It has been previously shown that controversial opinions exist
regarding the effects of the LOCOMAT-assisted gait training
in significantly improving the velocity and distance of the gait
as compared with the conventional one [16], [35]. Presumably,
this might be due to a lack of the motor recovery because of the
haptic guidance and biosensory feedback. According to a recent
experimental study, the robot-assisted gait training was more ef-
Fig. 4. Scores of each evaluation scale at baseline, four weeks, and eight weeks/ fective in achieving a recovery of motor functions as compared
with the conventional one during the flexion and extension of
There were significant differences in the FAC, BBS and K-MBI between
the two groups . But, there were no significant differences in the knee and ankle in normal healthy individuals [23]. Taken
mean scores of the EQ-5D and MAS between the two groups . together, our results suggest that the robot-assisted locomotor
Moreover, a post-hoc analysis showed that there were also significant training system might be effective in maximizing the degree of
differences in the FAC, BBS and K-MBI between the two groups the recovery when combined with the conventional one in pa-
.
tients with hemiparetic stroke.
There are several limitations of the current study as shown
with the conventional one as compared with the conventional below: 1) we enrolled a small number of patients in the current
one only. Most importantly, this is the first clinical evidence study. It is therefore difficult to generalize our results; 2) we en-
demonstrating superior therapeutic effects of the combination rolled the patients with sub-acute hemiparetic stroke. This sug-
of a conventional physical therapy and WALKBOT-assisted gests that clinicians should consider the effects of the robot-as-
gait training on balance, gait, and daily activity function. sisted training in making a spontaneous recovery of the func-
642 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 23, NO. 4, JULY 2015

tions when interpreting the results; 3) despite a lack of statistical [14] P. Liu et al., “Change of muscle architecture following body weight
significance, there was a difference in the ratio of ischemic-to- support treadmill training for persons after subacute stroke: Evidence
from ultrasonography,” Biomed. Res. Int., vol. 2014, p. 270676, 2014.
hemorrhagic stroke between the two groups. Further large-scale [15] J. H. Jung, N. G. Lee, J. H. You, and D. C. Lee, “Validity and feasibility
studies are therefore warranted to generalize our results. of intelligent Walkbot system,” Electron. Lett., vol. 45, pp. 1016–1017,
2009.
[16] T. G. Hornby et al., “Enhanced gait-related improvements after
therapist-versus robotic-assisted locomotor training in subjects with
V. CONCLUSION chronic stroke: A randomized controlled study,” Stroke, vol. 39, pp.
1786–1792, 2008.
[17] L. Gizzi et al., “Motor modules in robot-aided walking,” J. Neuroeng.
In conclusion, our results indicate that the WALKBOT might Rehabil., vol. 9, p. 76, 2012.
be a potentially more effective robotic-assisted gait training [18] C. Krishnan, R. Ranganathan, Y. Y. Dhaher, and W. Z. Rymer, “A pilot
system as compared with the conventional one in patients with study on the feasibility of robot-aided leg motor training to facilitate
active participation,” PLoS One, vol. 8, p. e77370, 2013.
hemiparetic stroke. Advanced research and development in [19] A. Mayr et al., “Prospective, blinded, randomized crossover study of
the robot-assisted locomotor training system will open up new gait rehabilitation in stroke patients using the Lokomat gait orthosis,”
possibilities and more opportunities for maximal restoration Neurorehabil. Neural Repair, vol. 21, pp. 307–314, 2007.
[20] A. Pennycott, D. Wyss, H. Vallery, V. Klamroth-Marganska, and R.
of the locomotor recovery and independent ambulation in Riener, “Towards more effective robotic gait training for stroke reha-
neurological populations with posture and gait impairment in bilitation: A review,” J. Neuroeng. Rehabil., vol. 9, p. 65, 2012.
the near future. But further large-scale studies are warranted to [21] J. Liepert et al., “Treatment-induced cortical reorganization after stroke
in humans,” Stroke, vol. 31, pp. 1210–1216, 2000.
establish our results. [22] E. C. Field-Fote and K. E. Roach, “Influence of a locomotor training
approach on walking speed and distance in people with chronic spinal
cord injury: A randomized clinical trial,” Phys. Ther., vol. 91, pp.
48–60, 2011.
REFERENCES [23] L. Marchal-Crespo, J. Schneider, L. Jaeger, and R. Riener, “Learning
a locomotor task: With or without errors?,” J. Neuroeng. Rehabil., vol.
[1] G. DeJong, S. D. Horn, B. Conroy, D. Nichols, and E. B. Healton, 11, p. 25, 2014.
“Opening the black box of post-stroke rehabilitation: Stroke rehabil- [24] H. S. Jorgensen, H. Nakayama, H. O. Raaschou, and T. S. Olsen, “Re-
itation patients, processes, and outcomes,” Arch. Phys. Med. Rehabil., covery of walking function in stroke patients: The Copenhagen Stroke
vol. 86, pp. S1–S7, 2005. Study,” Arch. Phys. Med. Rehabil., vol. 76, no. 1, pp. 27–32, 1995.
[2] A. S. Go et al., “Executive summary: Heart disease and stroke statis- [25] M. K. Holden, K. M. Gill, and M. R. Magliozzi, “Gait assessment for
tics—2014 update: A report from the American Heart Association,” neurologically impaired patients. Standards for outcome assessment,”
Circulation, vol. 129, pp. 399–410, 2014. Phys. Ther., vol. 66, pp. 1530–1539, 1986.
[3] E. Taub et al., “Technique to improve chronic motor deficit after [26] K. O. Berg, S. L. Wood-Dauphinee, J. L. Williams, and B. Maki, “Mea-
stroke,” Arch. Phys. Med. Rehabil., vol. 74, pp. 347–354, 1993. suring balance in the elderly: Validation of an instrument,” Can. J.
[4] S. H. You et al., “Virtual reality-induced cortical reorganization and as- Public Health, vol. 83, p. S7-1, 1992.
sociated locomotor recovery in chronic stroke: An experimenter-blind [27] F. M. Dias et al., “Functional capacity of oldest old living in a long-stay
randomized study,” Stroke, vol. 36, pp. 1166–1171, 2005. institution in Rio De Janeiro, Brazil,” J. Phys. Ther. Sci., vol. 26, pp.
[5] J. A. Kleim, S. Barbay, and R. J. Nudo, “Functional reorganization of 1097–1105, 2014.
the rat motor cortex following motor skill learning,” J. Neurophysiol., [28] F. M. Collen, D. T. Wade, and C. M. Bradshaw, “Mobility after stroke:
vol. 80, pp. 3321–3325, 1998. Reliability of measures of impairment and disability,” Int. Disabil.
[6] R. J. Nudo, B. M. Wise, F. SiFuentes, and G. W. Milliken, “Neural Stud., vol. 12, pp. 6–9, 1990.
substrates for the effects of rehabilitative training on motor recovery [29] J. M. Veerbeek et al., “What is the evidence for physical therapy post-
after ischemic infarct,” Science, vol. 272, pp. 1791–1794, 1996. stroke? A systematic review and meta-analysis,” PLoS One, vol. 9, no.
[7] C. E. Lang, J. R. MacDonald, and C. Gnip, “Counting repetitions: An 2, p. e87987, 2014.
observational study of outpatient therapy for people with hemiparesis [30] S. Grillner, “P. Central pattern generators for locomotion, with special
post-stroke,” J. Neurol. Phys. Ther., vol. 31, pp. 3–10, 2007. reference to vertebrates,” Annu. Rev. Neurosci., vol. 8, pp. 233–261,
[8] R. Riener et al., “Patient-cooperative strategies for robot-aided tread- 1985.
mill training: First experimental results,” IEEE Trans. Neural Syst. Re- [31] E. P. Zehr and J. Duysens, “Regulation of arm and leg movement
habil. Eng., vol. 13, no. 3, pp. 380–394, Sep. 2005. during human locomotion,” Neuroscientist, vol. 10, no. 4, pp. 347–361,
[9] P. Langhorne, F. Coupar, and A. Pollock, “Motor recovery after stroke: 2004.
A systematic review,” Lancet Neurol., vol. 8, pp. 741–754, 2009. [32] K. G. Pearson, “Common principles of motor control in vertebrates and
[10] J. M. Veerbeek et al., “What is the evidence for physical therapy post- invertebrates,” Annu. Rev. Neurosci., vol. 16, pp. 265–297, 1993.
stroke? A systematic review and meta-analysis,” PLoS One, vol. 9, p. [33] R. Riener et al., “Patient-cooperative strategies for robot-aided tread-
e87987, 2014. mill training: First experimental results,” IEEE Trans. Neural Syst. Re-
[11] K. W. Lau and M. K. Mak, “Speed-dependent treadmill training is habil. Eng., vol. 13, no. 3, pp. 380–394, Sep. 2005.
effective to improve gait and balance performance in patients with [34] C. E. Lang, J. R. MacDonald, and C. Gnip, “Counting repetitions: An
sub-acute stroke,” J. Rehabil. Med., vol. 43, pp. 709–713, 2011. observational study of outpatient therapy for people with hemiparesis
[12] Y. Laufer, R. Dickstein, Y. Chefez, and E. Marcovitz, “The effect of post-stroke,” J. Neurol. Phys. Ther., vol. 31, pp. 3–10, 2007.
treadmill training on the ambulation of stroke survivors in the early [35] J. Hidler, D. Nichols, M. Pelliccio, K. Brady, D. D. Campbell, and J.
stages of rehabilitation: A randomized study,” J. Rehabil. Res. Dev., H. Kahn et al., “Multicenter randomized clinical trial evaluating the
vol. 38, pp. 69–78, 2001. effectiveness of the Lokomat in subacute stroke,” Neurorehabil. Neural
[13] D. L. Damiano and S. L. DeJong, “A systematic review of the effec- Repair, vol. 23, pp. 5–13, 2009.
tiveness of treadmill training and body weight support in pediatric re-
habilitation,” J. Neurol. Phys. Ther., vol. 33, pp. 27–44, 2009. Authors’ photographs and biographies not available at the time of publication.