Hybrid Assistive Limb enhances the gait functions in sub-acute stroke stage: A multi single-case study

Abstract

BACKGROUND: Hybrid Assistive Limb (HAL) is a new robotics walking training device consisting of a wearable robot that interactively provides motion according to the wearer’s voluntary drive. The ability of HAL to enhance walking speed has not been clarified in sub-acute people with stroke, and few studies have focused on changes in walking pattern.

OBJECTIVES: To assess clinical availability of HAL in sub-acute stroke stage, we conduct a multi single-case study.

METHOD: An A-B-A single case study design was used in 4 patients. The intervention was conventional training in period A1 for 5 weeks and period A2 for 4 or 5 weeks and training with HAL and conventional training in period B for 5 weeks. The primary outcomes were maximum walking speed (MWS), cadence, and mean step length assessed each week. The secondary outcomes were Functional Ambulation Category, Berg Balance Scale, Fugl-Meyer Motor Assessment, and asymmetry rate in single-limb support time, as assessed before and after period B.

RESULTS: The significant increases in MWS was found in three patients through period B. The cadence, mean step length, FAC and AR were also increased in one, three, two and two patients, respectively.

CONCLUSIONS: It was found that HAL training of people with sub-acute stroke is an effective walking training to enhance the walking speed with the change of walking pattern in our clinical setting. Further studies are needed including control trials to analyze satirical difference.

Keywords

Hybrid Assistive Limb robotics walking training stroke

1 Introduction

Stroke is a major cause of walking disability worldwide [1]. More than 60% of people with stroke have walking disability and approximately 50% find it impossible to walk at disease onset [2]. Improving the walking ability of stroke is therefore an important goal, and various interventions have been tested in the past.

In 2003, Hesse et al. [3] pointed out that automated motor rehabilitation offers a fascinating new perspective on treatment, diagnosis, and interdisciplinary cooperation to the benefit of all participants, although a robot can never replace multi-level interaction between patients and therapist. Since then, research into the effects of robotic rehabilitation of the upper and lower limbs on central nervous disabilities has increased. A systematic review of electromechanical-assisted training for walking after stroke [4, 5] revealed that using electromechanical-assisted gait training devices in combination with physiotherapy increases the chance of regaining independent walking ability in people after stroke.

The Hybrid Assistive Limb (HAL) was developed by Cyberdyne Corporation (Tsukuba, Japan). HAL is a wearable robot that interactively provides motion according to the wearer’s voluntary drive [6]. HAL detects the bioelectrical signals generated by the patient’s muscle activity or the floor-reaction-force signals caused by the patient’s intended weight shifts, or both. HAL enables locomotor training with voluntary drive, and it has the advantages of both voluntary drive and ambulatory performance. Other exoskeletons use autonomously generated predefined motion. In contrast, HAL generates motion in response to the wearer’s voluntary drive. The wearer operates HAL by adjusting his or her muscle activity. Thus HAL is able to conduct locomotor training by providing motion support in response to the user’s voluntary drive. This assistance mechanism differs completely from those of other exoskeletons. In addition, other exoskeletons are designed for walking on a treadmill; therefore, they provide a simulated gait that differs from that of walking on a flat floor. In contrast, as a wearable system, HAL delivers locomotor training in an actual ambulatory environment. Kubota et al. [7] reported that the gait speed of patients with limited mobility, including people with chronic stroke, increases after gait training with HAL.

A recent randomized controlled trial comparing HAL gait training with conventional gait training showed that participants in the former group were significantly more able to walk independently after training [8]. Watanabe et al. remarked that the optimum duration of rehabilitation, the duration of each training session, the intervention frequency, and the long-term effects of HAL needed to be examined. In a previous unpublished study, we examined the effect of HAL training in people with sub-acute stroke. Motor recovery and walking function were greatly improved in all participants in the early recovery stage. However, improvement rates differed greatly among individuals. Large numbers of participants are therefore needed to detect any statistically significant effects of HAL training in a randomized controlled trial.

We therefore designed an intervention to occur at a time when improvement of walking speed was stable, in the late recovery stage of stroke, to clarify the effects of HAL walking training. We then tentatively evaluated the validity of this design in this pilot study. The first purpose of the study was to use a single case study design to explore the walking speed–enhancing effect of HAL gait training in people with sub-acute stroke whose walking speed had stabilized in the late recovery stage of their disease.

In addition, little has been reported on the effects of HAL gait training on cadence, step length (SL), and walking asymmetry in people with stroke. The second purpose of this study was therefore to examine changes in walking pattern and asymmetry after gait training with HAL.

2 Methods

2.1 Participants

Participants were recruited from an inpatient rehabilitation unit in Ibaraki Prefectural University of Health Sciences Hospital, Japan. All participants were admitted to our hospital between October 2013 and December 2013 through acute care hospitals to receive acute medical care and acute rehabilitation.

Inclusion criteria were diagnosis of first cerebral infarction or cerebral hemorrhage with hemiparesis and, for suitability for HAL, height from 150 to 185 cm and weight from 40 to 80 kg.

Exclusion criteria were high-risk heart disease, uncontrollable or severe high blood pressure, severe chronic respiratory disease, severe diabetes mellitus, severe aphasia, severe cognitive deficit lesion of the cerebellum or brain stem, subarachnoid hemorrhage, a need for severe risk control in physical therapy, severe sensory aphasia, ability to walk independently without cane or orthosis and live in the community, severe contracture and deformities of the lower limb, and use of an active implantable medical device.

The ethics committee of Ibaraki Prefectural University of Health Sciences approved the study, and written informed consent to participate was given by all participants or their legal representatives. This study was part of a research project, the protocol of which was registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN000012760).

2.2 Case description

Four patients participated in the study. Table 1 gives the participant profiles in terms of age, sex, weight, side of paresis, time from stroke to admission, and time from stroke to intervention. The first participant (Case 1) was post-left corona radiata infarction. Case 2 was post-right frontal lobe and parietal lobe hemorrhage. Case 3 was post-right putamen hemorrhage, and case 4 was post-right parietal lobe and capsula interna hemorrhage.

2.3 Design

An A-B-A design was used. Maximum walking speed was recorded during the baseline in period A (A1 = before B; A2 = after B). In period A, patients were treated by conventional physical therapy, which included gait training, muscle strength training, range of motion training, up- and downstairs training, and other types of individual training for 60 min 5 days a week for 5 weeks. In period B, patients were given HAL gait training for 20 min and conventional physical therapy for 40 min a day, 4 or 5 days a week (excluding Saturdays, Sundays, and national holidays) for 5 weeks. In all periods, patients were given occupational therapy and speech therapy as needed.

Because of the great degree of functional improvement in the subacute stage after stroke, it is difficult to discriminate between the results of one type of training and those of others. We therefore intervened at a time when the improvement in walking speed was stable. To determine when to start the intervention, we evaluated the patients’ walking speed by means of a 10-m walk test every week. The simple moving average (SMA) of walking speed in the preceding 3 weeks was then calculated. To assess the improvement rate, we calculated the change in the rate by dividing the difference between the present SMA and the previous week’s SMA by the previous SMA. We intervened at a time when the improvement in walking speed was stable. We defined this as occurring when the change in the rate was less than 10% in the first, second and third week and 5% in the fourth and fifth weeks consecutively. We then monitored whether or not the change in rate was stable for 5 consecutive weeks.

2.4 Intervention

The single-leg version of HAL was placed on the participant’s paretic side. To prevent falling, the HAL attached to the patient was connected to a mobile suspension system (All-In-One Walking Trainer, Ropox A/S, Næstved, Denmark) without the patient bearing weight (Fig. 1). During the intervention, gait training was implemented at a speed judged by the therapists to be as fast as possible while still maintaining a good gait posture on level ground. Gait training with HAL was done 4 or 5 times a week. One gait training session with HAL took a total of 20 min a day, excluding the device-fitting time and rest time.

During each training session, walking was continuously assisted by HAL, which is a computer-controlled exoskeletal device. The design and control system for HAL have been described in detail elsewhere [9–11, 30 , 9–11, 30]. The exoskeletal frame was fixed to the pelvis and the lower limbs by thigh and lower leg cuffs. Active actuators were installed at the hip joints and knee joints and generated assistive extension or flexion torque of these joints. HAL has two control systems, namely Cybernic Voluntary Control (CVC) mode and Cybernic Autonomous Control (CAC) mode [9]. CVC mode drives the amount and timing of the assistive torque provided to each joint for walking on the basis of bioelectrical signals from the flexor and extensor muscles at the hip and knee [12]. The gain in assistive torque at each joint in response to these bioelectrical signals was controlled by a therapist. The optimal gain and balance of the torque to maintain an appropriate walking pattern were determined by observing the joint trajectories and listening to the patient’s comments. If it is difficult for patients to achieve the motion derived from the CVC, the CAC can autonomously generate torque according to the walking pattern by referring to information from the floor reaction force [13]; CAC mode was thus used until the patient became familiar with the CVC.

2.5 Outcome measures

The primary outcomes were maximum walking speed (MWS) [14], cadence, and mean SL, using a 10-m walk test each week. If the participant’s Functional Ambulation Category (FAC) [15] was 2, then assistance in the 10-m walk test was needed only to prevent the participant from falling. In every test the physical therapist giving the assistance was different from the one who gave the conventional physical therapy. The evaluator, who was a third person, did not change from admission to discharge. If the FAC was 0 or 1, we waited until the FAC was 2 before we did the 10-m walk test. The time elapsed and number of steps were measured in the intermediate 6 m to allow for acceleration and deceleration. Cadence and mean SL [16] were calculated from the time elapsed and the number of steps in the 10-m section. The best MWS data from three trials were used for the analysis.

The secondary outcomes were the level of independence according to the FAC, the balance function according to the Berg Balance Scale (BBS) [17], motor recovery according to the Lower Extremity part of the Fugl-Meyer Motor Assessment (FMA-LE) [18] and the asymmetry ratio (AR) [19 –21] in swing time (equal to single-limb support time). Secondary outcomes were assessed on admission, at the start of period B, and at the end of period B. The evaluator was a different person from the therapist that gave the patient normal physical therapy. AR was evaluated at a self-selected walking speed by motion analysis. Gait motion was captured in the sagittal plane by a video camera (Sony HDR-CX 390, Tokyo, Japan; sampling rate 60 Hz). Motion times were applied to the gait after more than 5 steps from the starting line. The video times were loaded into a computer and used to calculate the single support time (length of time for which one leg was on the ground) in 5 walking cycles by using video editing software (EDIUS Neo 3.5, Grass Valley K.K., Japan).

2.6 Statistical analysis

We defined period A1 as the period of 5 weeks in which the change in improvement rate, as calculated by using the SMA, was less than 10% in the first week and 5% in the subsequent 4 weeks.

MWS, cadence, and SL were plotted on a graph weekly to enable visual analysis [22]. To detect changes in trends between periods A1, B, and A2, changes in plot level during a period, as well as the variability, direction, or slope on the graph, were needed [22]. Least-square lines were also drawn for visual analysis [23].

In addition, non-parametric statistical analyses were done to support the results observed in the graphs. Differences between each period (A1, B, A2) were inferred with the Kruskal-Wallis test.

Serial dependency was also assessed by using an autocorrelation coefficient [24]. Clinical ordinal consecutive data such as those in this report are not independent. Most of the data were autocorrelated (r > 0.20). Because autocorrelation increases theprobability of type I errors when there is a positive correlation [24], we chose a conservative P value of 0.01 [24, 25].

The differences in AR between the start and end of period B were analyzed by using a Wilcoxon signed-rank test. A P value less than 0.05 was considered statistically significant.

We used SPSS version 21.0 for all statisticalanalyses.

3 Results

All sessions were held for 20 min and completed safely. The number of sessions per participant ranged from 22 to 24 (Table 1).

Upon visual analysis, the graphs of weekly MWS showed improvement between periods A1 and B in Case 2, 3, and 4 (Fig. 2). The improvement was less clear in case 1. In period A2 the improvement decreased in case 1 and 2 and was constant in case 3 and 4.

The results for the graphs of cadence varied among individuals. There was evident improvement between periods A1 and B in case 3, whereas there was a clear decrease in case 1. There was a continued increase with a slightly smaller slope in period B compared with period A1 in case 2, and virtually no change in case 4 (Fig. 3). In case 1 and 3 the cadence in period A2 was greater than that in period B, whereas in case 4 the cadence was almost unchanged throughout.

The graphs of SL revealed obvious improvement between periods A1 and B in case 1, 2, and 4 (Fig. 4). The graph of SL for case 3 showed a decrease in improvement rate by visual analysis. In period A2 the improvement rate decreased in case 1 and 2 and increased in case 3 and 4.

The results of the non-parametric statistical analyses of MSW supported those of the visual analysis and were as follows: Case 1 to 4, respectively, Kruskal-Wallis χ2 = 8.68, P-value = 0.013; χ2 = 9.92, P-value = 0.0070; χ2 = 11.58, P-value = 0.0031; χ2 =11.57, P-value = 0.0031. The non-parametric statistical analyses for cadence also supported the visual analysis results and were as follows: χ2 = 7.43, P-value = 0.024; χ2 = 7.22, P-value = 0.027; χ2 = 12.02, P-value = 0.0025; χ2 = 4.71, P-value = 0.095. The non-parametric statistical analyses for SL supported the visual analysis results and were as follows: χ2 =11.79, P-value = 0.0028; χ2 = 10.14, P-value = 0.0063; χ2 = 10.44, P-value = 0.0054; χ2 = 12.69, P-value =0.0018.

Table 1 shows the differences in secondary outcomes between the start and the end of period B. FAC improved in case 1 and 2 during period B. FAC in case 4 was 5 at the start of period B. BBS and FMA-LE improved in all participants during period B. AR decreased in all participants and changed significantly in case 1 and 2 (P < 0.05).

4 Discussion

In Japan, it is usual for people with sub-acute stroke that are considered to require more rehabilitation to be transferred to a recovery rehabilitation hospital or unit. In our unpublished pilot study we intervened in the early recovery stage. However, it was difficult to extract the effect of only the HAL training from our results.

Therefore, in this current pilot study we explored the effect of HAL gait training in enhancing walking speed in people with stroke in the late recovery stage. The MWS results clearly showed improvement in period B in case 2, 3, and 4 and a slight improvement in case 1, whereas the participants’ MWS values were almost stable in period A1. All participants accordingly started receiving their interventions at more than 100 days. Therefore, we consider that our design, whereby we intervened with HAL when the improvement in walking speed was stable, was reasonable.

Improvements in MWS were caused by an increase in SL in case 1, 2, and 4 and by an increase in cadence in case 3. A recent randomized controlled trial byWatanabe et al. [8] showed a significant difference between a HAL training group and a group receiving conventional training, but only in FAC. We found here that MWS was better in period B than in period A1 in the majority of patients. This difference between the studies may have been caused by the fact that we intervened for longer and more often than in the trial by Watanabe et al., who used HAL for 4 weeks at 3 times a week. However, our number of participants was only small.

In addition, the improvements in MWS in period B were the result of not only an increase in cadence or SL, but also a decrease in the other parameter. For example, in case 1, cadence decreased and SL increased in period B, conversely in case 3 cadence increased and SL decreased in period B. The difference in SL between the HAL group and the conventional group was not significant in the trials performed by Watanabe et al. However, we found here that the increase or decrease in SL or cadence depended on the individual. In healthy adults, the walk ratio (calculated by dividing SL by cadence) [26] is generally constant; however, because walking speed can be changed, it is affected by SL and cadence speed to the same extent [27, 28]. Furthermore, Suzuki et al. [29] indicated that people with hemiparetic stroke in the recovery stage improved in MWS, with an invariant relationship between stride length and walking rate. Our results suggest that HAL training changed the participants’ gait patterns. In fact, our finding that AR decreased in all participants and significantly increased in case 1 and 2 showed that gait symmetry improved. The ability to maintain single-limb support is an important determinant of gait stability [30, 31]. Asymmetry during single-limb support seems to be related to decreased ability to bear weight on the paretic limb [32], and single-support training helps to achieve symmetric gait in people with stroke [33, 34]. Goldie et al. [35] also pointed that increase in single support time on the paretic side is a good indicator of increase in weight bearing on the paretic side, whereas increase in single support time on the non-paretic side is a good indicator of better paretic leg advancement. In addition, training that incorporates active participation whereby the patient voluntarily produces movement, inducing changes in motor performance, cortical activity, and excitability, is considered essential for motor learning of new tasks [36, 37]. Therefore, training with HAL, which interactively provides motion according to the wearer’s voluntary drive, may have enhanced walking performance in association with an improvement in walking pattern.

FAC was improved in case 1 and 3. Case 4 already had a perfect FAC score of 5 at the start of period B. BBS in all participants and FMA-LE in 3 participants (excluding case 2) increased; these changes were similar to those in the trial by Watanabe et al. [8].

A limitation of our study was the small number of participants. However, as a feasibility study its aim was to explore the effects of HAL training on walking speed at a time when the improvement in walking speed had stabilized in the late recovery stage, as well as to explore changes in walking pattern and asymmetry. Further studies are needed to explain how HAL training improves walking speed and walking pattern. We intend to use more participants and create a control group that will be treated only with conventional physical therapy and not with HAL training.

The results of this exploratory study suggested that HAL training in people with sub-acute stroke may enhance walking speed in association with an improvement in walking pattern at a time when the rate of improvement in walking speed is stable in the late recovery stage. Further studies with a control group are needed to clarify the effects on walking function.

Conflict of interest

Kenichi Yoshikawa, Masafumi Mizukami, Ayumu Sano, Kazunori Koseki, Yuko Hashizume, Yasutsugu Asakawa, Yutaka Kohno, Hiroshi Nagata, Kei Nakai, and Hideo Tsurushima have no competing interests to declare. Hiroaki Kawamoto is a founder, shareholder, and external director of CYBERDYNE Inc., which produces HAL.

Contributors

KY participated in the design of the study, carried out the data collection, analysis and interpretation, drafted and revised the manuscript.

MM participated in the design of the study, collected patient’s informed consent and drafted the manuscript.

AS and KK carried out the HAL training.

YH and KI carried out the assessments.

YA participated in the statistical analysis.

HK participated in the design of the study and provided the HAL suits and technical support.

YK, HN, KN and HT participated in the coordination and design of the study, in finalizing the manuscript.

Funding

This work was supported by Grant-in-Aid for Project Research (1247) from Ibaraki Prefectural University of Health Sciences.

Footnotes

Acknowledgments

We thank colleagues in Dept. of Physical Therapy, Ibaraki Prefectural University of Health Sciences Hospital and physical therapists in for their help with this study. The authors would particularly like to thank the study participants for their time.

References

Feigin

, Lawes

, Bennett

, Barker-Collo

, Parag

. Worldwide stroke incidence and early case fatality reported in 56 population-based studies: A systematic review. The Lancet Neurology 2009;8(4):355–369.

Jørgensen

, Nakayama

, Raaschou

, Vive-Larsen

, Støier

, Olsen

. Outcome and time course of recovery in stroke. Part II: Time course of recovery. The Copenhagen Stroke Study. Archives of Physical medicine and Rehabilitation 1995;76(5):406–412.

Hesse

, Schmidt

, Werner

, Bardeleben

. Upper and lower extremity robotic devices for rehabilitation and for studying motor control. Current Opinion in Neurology 2003;16(6):705–710.

Mehrholz

, Elsner

, Werner

, Kugler

, Pohl

. Electromechanical-assisted training for walking after stroke. The Cochrane Database of Systematic Reviews 2013;7, Cd006185.

Mehrholz

, Elsner

, Werner

, Kugler

, Pohl

. Electromechanical-Assisted Training for Walking After Stroke Updated Evidence. Stroke 2013;44(10):e127–e128.

Kawamoto

, Kamibayashi

, Nakata

, Yamawaki

, Ariyasu

, Sankai

, et al. Pilot study of locomotion improvement using hybrid assistive limb in chronic stroke patients. BMC Neurology 2013;13, 141.

Kubota

, Nakata

, Eguchi

, Kawamoto

, Kamibayashi

, Sakane

, et al. Feasibility of rehabilitation training with a newly developed wearable robot for patients with limited mobility. Archives of Physical Medicine and Rehabilitation 2013;94(6):1080–1087.

Watanabe

, Tanaka

, Inuta

, Saitou

, Yanagi

. Locomotion Improvement Using a Hybrid Assistive Limb in Recovery Phase Stroke Patients: A Randomized Controlled Pilot Study. Archives of Physical Medicine and Rehabilitation 2014.

Suzuki

, Mito

, Kawamoto

, Hasegawa

, Sankai

. Intention-based walking support for paraplegia patients with Robot Suit HAL. Advanced Robotics 2007;21(12):1441–1469.

10.

Kawamoto

, Taal

, Niniss

, Hayashi

, Kamibayashi

, Eguchi

, et al. editors. Voluntary motion support control of Robot Suit HAL triggered by bioelectrical signal for hemiplegia. Engineering in Medicine and Biology Society (EMBC), Annual International Conference of the IEEE; 2010; IEEE.

11.

Kawamoto

, Sankai

. Power assist system HAL-3 for gait disorder person. Computers helping people with special needs 2002;19–29.

12.

Kawamoto

, Sankai

. Power assist system HAL-3 for gait disorder person. Computers helping people with special needs: Springer; 2002;196–203.

13.

Kawamoto

, Sankai

. Power assist method based on phase sequence and muscle force condition for HAL. Advanced Robotics 2005;19(7):717–34.

14.

Suzuki

, Imada

, Iwaya

, Handa

, Kurogo

. Determinants and predictors of the maximum walking speed during computer-assisted gait training in hemiparetic stroke patients. Archives of Physical Medicine and Rehabilitation 1999;80(2):179–82.

15.

Raphael

, Briscoe

, Davies

, Ian Whinnett

, Manisty

, Sutton

, et al. Limitations of the New York Heart Association functional classification system and self-reported walking distances in chronic heart failure. Heart (British Cardiac Society) 2007;93(4):476–82.

16.

Aaslund

, Helbostad

, Moe-Nilssen

. Walking during body-weight-supported treadmill training and acute responses to varying walking speed and body-weight support in ambulatory patients post-stroke. Physiotherapy Theory and Practice 2013;29(4):278–289.

17.

Berg

, Wood-Dauphinee

, Williams

. The Balance Scale: Reliability assessment with elderly residents and patients with an acute stroke. Scandinavian Journal of Rehabilitation Medicine 1995;27(1):27–36.

18.

Fugl-Meyer

, Jaasko

, Leyman

, Olsson

, Steglind

. The post-stroke hemiplegic patient. 1. a method for evaluation of physical performance. Scandinavian Journal of Rehabilitation Medicine 1975;7(1):13–31.

19.

Patterson

, Gage

, Brooks

, Black

, McIlroy

. Evaluation of gait symmetry after stroke: A comparison of current methods and recommendations for standardization. Gait & Posture 2010;31(2):241–246.

20.

Patterson

, Gage

, Brooks

, Black

, McIlroy

. Changes in gait symmetry and velocity after stroke: A cross-sectional study from weeks to years after stroke. Neurorehabilitation and Neural Repair 2010;24(9):783–790.

21.

Lewek

, Bradley

, Wutzke

, Zinder

. The relationship between spatiotemporal gait asymmetry and balance in individuals with chronic stroke. Journal of Applied Biomechanics 2014;30(1):31–36.

22.

Kratochwill

, Hitchcock

, Horner

, Levin

, Odom

, Rindskopf

, et al. Single-case designs technical documentation. What Works Clearinghouse, 2010.

23.

Johnson

, Ottenbacher

. Trend line influence on visual analysis of single-subject data in rehabilitation research. Disability & Rehabilitation 1991;13(2):55–59.

24.

Franklin

, Allison

, Gorman

. Design and analysis of single-case research, Psychology Press, 2014.

25.

Sideridis

, Greenwood

. Is human behavior autocorrelated? An empirical analysis. Journal of Behavioral Education 1997;7(3):273–293.

26.

Sekiya

, Nagasaki

. Reproducibility of the walking patterns of normal young adults: Test-retest reliability of the walk ratio (step-length/step-rate). Gait & Posture 1998;7(3):225–227.

27.

Sekiya

, Nagasaki

, Ito

, Furuna

. The invariant relationship between step length and step rate during free walking. Journal of Human Movement Studies 1996;30(6):241–257.

28.

Zijlstra

, Rutgers

, Hof

, Van Weerden

. Voluntary and involuntary adaptation of walking to temporal and spatial constraints. Gait & Posture 1995;3(1):13–18.

29.

Suzuki

, Yamada

, Handa

, Imada

, Iwaya

, Nakamura

. Relationship Between Stride Length and Walking Rate in Gait Training for Hemiparetic Stroke Patients1. American Journal of Physical Medicine & Rehabilitation 1999;78(2):147–152.

30.

Perry

, Davids

. Gait analysis: Normal and pathological function. Journal of Pediatric Orthopaedics 1992;12(6):815.

31.

Hendrickson

, Patterson

, Inness

, McIlroy

, Mansfield

. Relationship between asymmetry of quiet standing balance control and walking post-stroke. Gait & Posture 2014;39(1):177–181.

32.

Hsu

, Tang

, Jan

. Analysis of impairments influencing gait velocity and asymmetry of hemiplegic patients after mild to moderate stroke. Archives of Physical Medicine and Rehabilitation 2003;84(8):1185–1193.

33.

Olney

, Richards

. Hemiparetic gait following stroke. Part I: Characteristics. Gait & Posture 1996;4(2):136–148.

34.

Aruin

, Rao

, Sharma

, Chaudhuri

. Compelled body weight shift approach in rehabilitation of individuals with chronic stroke. Topics in Stroke Rehabilitation 2012;19(6):556–563.

35.

Goldie

, Matyas

, Evans

. Gait after stroke: Initial deficit and changes in temporal patterns for each gait phase. Archives of Physical Medicine and Rehabilitation 2001;82(8):1057–1065.

36.

Lotze

, Braun

, Birbaumer

, Anders

, Cohen

. Motor learning elicited by voluntary drive. Brain 2003;126(4):866–872.

37.

Pennycott

, Wyss

, Vallery

, Klamroth-Marganska

, Riener

. Towards more effective robotic gait training for stroke rehabilitation: A review. Journal of Neuroengineering and Rehabilitation 2012;9, 65.