Comparison of active-assisted and active-unassisted robot-mediated upper limb therapy in subacute stroke

Abstract

Background:

Upper-limb robot-mediated therapy is usually carried out in active-assisted mode because it enables performance of many movements. However, assistance may reduce the patient’s own efforts which could limit motor recovery.

Objective:

The aim of this study was to compare the effects of active-assisted and active-unassisted robotic interactions on motor recovery in subacute stroke patients with moderate hemiparesis.

Methods:

Fourteen patients underwent a 6-week combined upper limb program of usual therapy and robotic therapy using either the active-unassisted (n = 8) or active-assisted (n = 6) modes. In the active-assisted group, assistance was only provided for the first 3 weeks (1st period) and was then switched off for the remaining 3 weeks (2nd period). The Fugl-Meyer Assessment (FMA) was carried out pre- and post-treatment. The mean number of movements performed and the mean working distance during the 1st and 2nd periods were compared between groups.

Results:

FMA score improved post-treatment in both groups with no between-group differences: active-assisted group: +8±6 pts vs active-unassisted group: +10±6 pts (ns). Between the 1st and 2nd periods, there was a statistical trend towards an improvement in the number of movements performed (p = 0.06) in the active-unassisted group (526±253 to 783±434, p = 0.06) but not in the active-assisted group (882±211 to 880±297, ns). Another trend of improvement was found for the working distance in the active-unassisted group (8.7±4.5 to 9.9±4.7, p = 0.09) but not in the active-assisted group (14.0±0 to 13.5±1.1, ns).

Conclusions:

The superiority of the non-assistive over assistive robotic modes has not been demonstrated. However, the non-assistive mode did not appear to reduce motor recovery in this population, despite the performance of fewer movements on shorter working distance compared with the group who had assistance. It seems that the requirement of effort could be a determinant factor for recovery in neurorehabilitation however further well-design studies are needed to fully understand this phenomenon.

Keywords

Robot-mediated therapy active-unassisted mode upper limb stroke.

1 Introduction

Following stroke, approximately 80% of patients are left with persistent upper limb functional limitations that impact on their capacity to carry out activities of daily living (Nichols-Larsen et al., 2005). In order to promote activity-dependent plasticity and to optimize the recovery of motor function, neurorehabilitation needs to encompass motor learning principles including repetitive active movement-based training (Krakauer, 2006). Robot-mediated therapy provides these intensity-related neuroplastic factors, by increasing the therapy time (robot-mediated therapy in addition to usual therapy), and by varying the interaction modes (active, active-assisted, constrained). Upper limb robot-assisted training integrated into a post-stroke rehabilitation program reduces motor impairment in patients with moderate to severe impairment in both the subacute and chronic phases of stroke (Mehrholz et al., 2018). However the translation of this reduction of motor impairment into improved upper limb function is not clear; the RATULS trial was expected to clarify the issue with the enrollment of 770 patients with moderate to severe upper limb paresis after stroke (Rodgers et al., 2019). Unfortunately, while this large study (probably one of the largest randomized controlled trials of stroke rehabilitation) confirmed that robotic training improved motor outcomes, it did not result in greater functional improvement than enhanced upper limb therapy. In robot-mediated therapy, more than 1000 movements are performed by patients in one session compared to 50 movements in usual therapy (Lang et al., 2009). Direct comparison with intensity-matched usual therapy (same therapy time and number of movements) showed similar improvements in motor function (Lo et al., 2010), however, it is unrealistic for patients to perform such high numbers of movement repetitions routinely in usual therapy. In robot-mediated therapy, the active-assisted mode facilitates the performance of many movements since the robot assists the movement. However, one disadvantage of the provision of assistance is that it could lead to a reduction in the effort made to produce movement (Marchal-Crespo & Reinkensmeyer, 2009), as seen in the lower limb (Miyai et al., 2006). To minimize slacking and increase patient engagement, novel classes of robotic devices and new therapeutic approaches combining robot-mediated therapy with a motor learning task have been developed and shown to be effective (Krishnan et al., 2012; Krishnan, Ranganathan, et al., 2013; Washabaugh et al., 2018, 2019; Washabaugh & Krishnan, 2018). In addition, new control algorithms have been developed (Krishnan, Kotsapouikis, et al., 2013) in which the assist-as-needed concept has been optimized (Emken et al., 2005; Reinkensmeyer et al., 2007). However few studies have explored the effect of the robotic assistance itself on motor impairments in hemiparesis. The assist-as-needed mode has been shown to help to sustain engagement during hundreds of movement repetitions in patients with severe paresis (Grosmaire & Duret, 2017), however the question of the utility of assistance has not been addressed in patients with moderate paresis. The aim of this study was therefore to compare the effects of active-assisted and active-unassisted robotic therapy on upper limb motor outcomes in patients with moderate paresis in the subacute phase of stroke. The study hypothesis was that patients with moderate paresis would reduce their efforts when using the assisted-as-needed mode, and that this would lead to a smaller degree of motor recovery.

2 Material and methods

2.1 Patients

The cohort used in this study was a convenience sample drawn from a larger cohort of inpatients who attended the Neurorehabilitation Department of the “Les Trois Soleils” Center (Boisisse-Le-Roi, France). All had previously participated in a prospective clinical trial (reference number: ID RCB 2011-A00632-39) that had been approved by the local Ethics Committee at Paris V and was conducted in accordance with both the Declaration of Helsinki (2008) and local regulatory requirements. All the patients (including those whose data were used in the present study) met the inclusion criteria of the original clinical trial i.e. (i) aged over 18; (ii) hemiparesis due to a first unilateral stroke that occurred between 4 and 10 weeks before inclusion in the study; (iii) active shoulder flexion on the paretic side ≥15°. All patients provided written informed consent regarding their participation in the original clinical trial. Patients were previously randomly allocated to one of two groups: active-assisted and active-unassisted. All subjects received the planned treatment and all were included in the statistical analysis (Fig. 1)

Fig. 1

Experimental protocol

2.2 Interventions

All patients participated in an upper-limb rehabilitation program involving 30 minutes of robot-mediated therapy and 30 minutes of usual therapy, five days a week for six weeks. Usual therapy included passive muscle stretching, active reaching movements and specific grasp and release tasks. Robot-mediated therapy was carried out with the InMotion Arm^TM robotic system (Interactive Motion Technologies, Inc, Watertown, MA, Fig. 2-A) (Krebs et al., 1998) which is a two-degree-of-freedom end-effector that trains shoulder and elbow movements in the horizontal plane. Patients were seated in front of a screen with their trunk constrained by a strap. They held the manipulandum with their paretic hand while their forearm was supported on a platform. The therapy consisted of a pointing task that involved repeated center-out movements (and back to center) towards eight visual targets located around a circle on the screen (Fig. 2-B).

Fig. 2

InMotion 2.0 shoulder/elbow robotic system. A: Therapist using the InMotion 2.0 shoulder/elbow robotic system; B: Pointing task interface.

Patients in the active-assisted group performed movements with robotic assistance for the first three weeks (1st period) and movements without assistance (assistance was switched off) for the remaining three weeks (2nd period). Patients in the active-unassisted group performed unassisted movements for the whole six weeks, but before each center-out movement the therapist could position the end-effector in the center of the task (if necessary) to encourage the patient to initiate movements toward the targets. Four center-out distances were available (3, 5, 10 and 14 cm) and the distance was chosen by the therapist according to the patient’s motor ability.

2.3 Clinical and robot-based assessments

Patients were assessed at pre- and post-treatment using the Fugl-Meyer Assessment for the upper extremity (FMA) (Gladstone et al., 2002). The FMA is widely used in studies of robotic rehabilitation and is a reliable, validated and responsive measure of upper-limb impairment in individuals with subacute stroke (Sivan et al., 2011). The scale, with a maximum score of 66 points, measures the capacity of the shoulder, elbow, wrist and hand to carry out selective movements. A kinematic measure, reaching distance (cm), defined as the mean curvilinear distance travelled by the hand from the starting position towards the target, was computed from 80 non-assisted center-out and out-center pointing movements in eight directions located around a 14 cm-radius circle. The raw data were filtered using a tenth-order Butterworth low-pass filter at a cut-off frequency of 5 Hz (Mazzoleni et al., 2013). In addition, during the robot sessions, two parameters were recorded for both groups and were averaged separately for the 1st period (average of all robotic sessions performed during the first 3 weeks) and the 2nd period (average of all robotic sessions performed during the remaining 3 weeks): the mean number of movements performed and the mean working distance in cm, based on the center-out distance chosen by the therapist for each session.

3 Statistical analysis

Pre-treatment between-group differences in pa-tient characteristics (age, time since stroke at the start of the protocol, FMA score and reaching distance) were analyzed using a Mann Whitney test. Pre/post changes in FMA score, pre/post changes in reaching distance and changes in the number of movements and working distance between the 1st and 2nd periods were compared between groups using a Mann-Whitney test. Within-group differences in the number of movements and working distance between the 1st and 2nd periods were analyzed using a Wilcoxon test. The p-value was set at 0.05 for all statistical analyses.

4 Results

Fourteen patients were included in the present study. Their demographic characteristics are reported in Table 1. At pre-treatment, only the reaching dis-tance differed between groups (active-assisted group vs active-unassisted group, p < 0.05); it was higher in the active-assisted group. There was no between-group difference in the pre/post change in FMA score (Fig. 3-A): +8±6 points (mean±SD) in the active-assisted group and +10±6 points in active-unassisted group (active-assisted group vs active-unassisted group, ns). There was no between-group difference in pre/post change in reaching distance: +2.7±3.4 cm in the active-assisted group and +4.1±3.8 cm in the active-unassisted group (active-assisted group vs active-unassisted group, ns; Fig. 3-B). The results for the mean number of movements and for the mean working distance are presented in Fig. 4. Between the 1st period and the 2nd period, the changes were marked by a statistical trend towards between-group differences for the mean number of movements (active-assisted group vs active-unassisted group: – 2±131 vs +257±240, p = 0.06) and for the mean working distance (active-assisted group vs active-unassisted group: – 0.5±1.1 cm vs +1.1±1.6 cm, p = 0.06). The within-group comparison between the 1st period and the 2nd period showed that in the active-unassisted group, both the mean number of movements (+42%; 1st period vs 2nd period, p = 0.06) and the mean working distance (+17%; 1st period vs 2nd period, p = 0.09) tended to increase. Those variables did not change in the active-assisted group (mean number of movements: – 2%, 1st period vs 2nd period, ns and mean working distance: – 3%, 1st period vs 2nd period, ns).

Table 1
Baseline patient characteristics

All Active-assisted group Active-unassisted group

Number of participants 14 6 8

Age (years) 66±10 67±11 65±10

Sex 5 F, 9 M 1 F, 5 M 4 F, 4 M

Side of paresis 9 R, 5 L 5 R, 1 L 4 R, 4 L

Stroke type 12 I, 2 H 4 I, 2 H 8 I, 0 H

Time since stroke at the start of the protocol (days) 56±17 49±12 62±18

Fugl-Meyer Assessment (/66) 32±15 40±14 26±13

Reaching distance (cm) 7.3±4.8 10.7±3.2 4.8±4.3^*

	All	Active-assisted group	Active-unassisted group
Number of participants	14	6	8
Age (years)	66±10	67±11	65±10
Sex	5 F, 9 M	1 F, 5 M	4 F, 4 M
Side of paresis	9 R, 5 L	5 R, 1 L	4 R, 4 L
Stroke type	12 I, 2 H	4 I, 2 H	8 I, 0 H
Time since stroke at the start of the protocol (days)	56±17	49±12	62±18
Fugl-Meyer Assessment (/66)	32±15	40±14	26±13
Reaching distance (cm)	7.3±4.8	10.7±3.2	4.8±4.3^*

Results are expressed as means±SD. F, female; M, male; R, right; L, left; I, ischemia; H, haemorrhage. ^*p < 0.05, Mann Whitney test, Active-assisted group vs. Active-unassisted group.

Fig. 3

Clinical and kinematic outcomes. Results are expressed as means (SEM).

Fig. 4

Robot-based outcomes. Results are expressed as means (SEM).

5 Discussion

The present study compared the effects of active-assisted and active-unassisted robotic therapy on clinical and robot-performance variables during a 6-week upper limb rehabilitation program in patients with moderate paresis in the subacute phase of stroke. Although the reduction in impairment (FMA score) was similar in both groups post-treatment, there was a tendency towards an improvement in the two robot-based variables (number of movements and working distance) between the first and second periods only in the active-unassisted group.

The lack of difference in the two robot-based parameters between the 1st and 2nd periods in the active-assisted group might be due to a ceiling effect (due to the movement assistance), associated with the effect of changing the mode of interaction of the robot therapy. Indeed, performance might have been biased by the effect of the movement assistance itself, which in this case, forced the patients to reach the maximal center-out distance of the task (i.e. 14 cm, see Fig. 4). Moreover, this assistance positively influenced the number of movements performed because when the patient cannot reach the target, the robot will complete the movement to validate the target. Furthermore, the change in interaction mode at mid-treatment by turning the assistance off could have masked the effort provided by these patients between both periods. As the performance of patients in the 1st period do not reflect their actual performance, the comparisons between both periods after changing the interaction mode will be biased. Another point of note is that the pointing task used in the robot-mediated therapy may not have challenged these patients with mild-moderate paresis. At the pre-treatment evaluation, the patients in the active-assisted group had a reaching distance (without assistance) of 10 cm (out of 14 cm) compared to the unassisted group who had a reaching distance of only around 5 cm. The implications of this are two-fold: 1) the patients in active-assisted group had a lower potential of improvement on the requested task and 2) the patients in the active-unassisted group had more severe initial impairments, which may have made recovery more difficult. Indeed, according to the classification by Duncan et al., based on FMA score, patients in the active-unassisted group would be classified as having moderate-severe impairment while the active-assisted group would be classified as having mild-moderate impairment (Duncan et al., 2000). While this difference in baseline impairment limits the comparison between groups, it is nevertheless of interest in the interpretation of the findings. We would have expected a decline in performance in robotic training without assistance to movement in the more severely impaired active-unassisted group between the 1st and 2nd periods (robotic task would be too difficult and patients would fail on this task). Instead, the mean number of movements and mean working distance improved, showing not only that patients do not fail but also succeed. This finding could support the postulate of Khan et al. (Kahn et al., 2001) that the key stimulus for motor recovery is repetitive attempts of active movements rather the performance of a large number of assisted movements. The results of the present study suggest that the repeated effort made by the patients (i.e. difficulty) may have been a stronger stimulus for activity-dependent plasticity. Since the performance of unassisted movements requires maximal effort (Miyai et al., 2006), it is likely that the patients in the active-unassisted group generated more effort to produce movement than the patients in the active-assisted group during the rehabilitation sessions. Overall, the results of this retrospective study suggest that the assist-as-needed robot mode is not adapted to all levels of motor impairment in patients with stroke. Future studies should aim to specify the conditions for which the use of assistance in robot-mediated therapy may be necessary and/or beneficial. A previous study of conventional therapy showed that it was feasible for patients in the subacute phase of stroke to perform a mean number of 289 movements during sessions of 47 to 60 minutes duration, but this was associated with a high level of fatigue and a significant increase in pain scores (Waddell et al., 2014). In the present study, the patients performed more than 500 unassisted movements in 30 minutes, illustrating the good tolerance and feasibility of a high dose of active-unassisted robotic therapy in patients with moderate paresis in the subacute phase of stroke. Further studies of fatigability, pain, energy consumed, attention and motivation are now required to objectively quantify the impact of these factors on the motor recovery following stroke.

The findings of this study also suggest that the choice of the mode of robot-therapy may im-pact on the neuro-plastic changes that accompany movement practice. The results of the recent Robot-Assisted Training for the Upper Limb after Stroke (RATULS) randomized controlled trial (n = 770), show that robot-mediated upper limb therapy did not improve upper limb function more than usual therapy (Rodgers et al., 2019). Although these results may be discouraging, it should be noted that robotic therapy is not intended to improve functional abilities. In fact, robotic therapy provides an impairment-based treatment from a bottom-up approach that consists of concatenating the functional task of several primary movements, leaving the functional integration of these gains for a later phase (Krebs et al., 2008).

However, the results of this study should be interpreted with caution since the study has several limitations. Firstly, the very small sample size, the associated lack of statistical power, the lack of blinding, and the retrospective nature of this study strongly limit the interpretation of the data. The cohort presented here was a convenience sample from a previous randomized prospective clinical trial, the results of which will subsequently be communicated. The larger study was not designed to compare active-assisted and unassisted modes. However, in view of the importance of that question, we took the opportunity to use some data available to provide preliminary answers. A second limitation is the atypical study design in which one group (active-assisted) participated in a crossover design (i.e., performed active-assisted mode for first 3 weeks and then performed active-unassisted mode for the next 3 weeks) while the other group (active-unassisted) performed unassisted robotic therapy for the entire 6 weeks. This experimental protocol is clearly inappropriate to analyze the effect of the assistance to movement on motor recovery. However, it could be interesting to use such a crossover design in clinical practice to challenge patients by grading the difficulty of the task for those with severe-to-moderate paresis. By turning the assistance off, they are forced to produce greater efforts, resulting in more intensive training, which is a potential determining factor of activity-dependent plasticity. Thirdly, since this study was performed in individuals who were in the subacute phase of stroke, the improvements that occurred over time were likely to have been due to the combined effect of lesion-induced plasticity and behavioral-induced plasticity. Finally, according to the reaching distance and FMA scores, the active-unassisted group had more severe impairment than the active-assisted group at baseline, thus all differences must be interpreted with caution.

6 Conclusions

This study is a first step in the analysis of the effects of assistance to movement in robotic therapy on motor recovery in patients with moderate paresis in the subacute phase of stroke. Overall, the results of this study did not provide a sufficient level of evidence of the superiority of the non-assistive mode over assistive mode in robotic therapy. Despite this, the non-assistive mode seems not to be detrimental on motor recovery in subacute stroke patients with moderate paresis suggesting that the requirement of effort could be a determinant factor for recovery in neurorehabilitation. Further well-designed studies with appropriate sample sizes are needed to fully understand the effects of the active-unassisted mode of robotic therapy for stroke rehabilitation.

Funding source

We would like to thank A.D.I.R.R (Association for Development and Innovation in Rehabilitation Robotics), an independent French association in providing a financial support for the overall preparation of the article (data collection, statistical analysis, interpretation of the data and writing of the report).

Declarations of interest

We have no conflict of interest.

Footnotes

Acknowledgments

We would like to thank the therapists at “Les Trois Soleils” rehabilitation center for their daily involvement in the robot-assisted program, without which this study would not have been possible. We would also like to thank Johanna Robertson and Jennifer Dandrea Palethorpe for English language improvement.

References

Duncan,

P.W.

, Min Lai,

, & Keighley,

(2000). Defining post-stroke recovery: Implications for design and interpretation of drug trials. Neuropharmacology, 39(5), 835–841.

Emken,

J.L.

, Bobrow,

J.E.

, & Reinkensmeyer,

D.J.

(2005). Robotic movement training as an optimization problem: designing a controller that assists only as needed. 9th International Conference on Rehabilitation Robotics, 2005. ICORR 2005., 307-312.

Gladstone,

D.J.

, Danells,

C.J.

, & Black,

S.E.

(2002). The Fugl-Meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabilitation and Neural Repair, 16(3), 232–240.

Grosmaire,

A.G.

, & Duret,

(2017). Does assist-as-needed upper limb robotic therapy promote participation in repetitive activity-based motor training in sub-acute stroke patients with severe paresis? NeuroRehabilitation, 41(1), 31–39.

Kahn,

L.E.

, Averbuch,

, Rymer,

W.Z.

, & Reinkensmeyer,

D.J.

(2001). Comparison of robot-assisted reaching to free reaching in promoting recovery from chronic stroke. Proceedings of the International Conference on Rehabilitation Robotics, 39-44.

Krakauer,

J.W.

(2006). Motor learning: its relevance to stroke recovery and neurorehabilitation. Current Opinion in Neurology, 19(1), 84–90.

Krebs,

H.I.

, Hogan,

, Aisen,

M.L.

, & Volpe,

B.T.

(1998). Robot-aided neurorehabilitation. IEEE Transactions on Rehabilitation Engineering, 6(1), 75–87.

Krebs,

H.I.

, Mernoff,

, Fasoli,

S.E.

, Hughes,

, Stein,

, & Hogan,

(2008). A Comparison of functional and impairment-based robotic training in severe to moderate chronic stroke: a pilot study. NeuroRehabilitation, 23(1), 81–87.

Krishnan,

, Kotsapouikis,

, Dhaher,

Y.Y.

, & Rymer,

W.Z.

(2013). Reducing robotic guidance during robot-assisted gait training improves gait function: A case report on a stroke survivor. Archives of Physical Medicine and Rehabilitation, 94(6), 1202–1206.

10.

Krishnan,

, Ranganathan,

, Dhaher,

Y.Y.

, & Rymer,

W.Z.

(2013). A pilot study on the feasibility of robot-aided leg motor training to facilitate active participation. PloS One, 8(10).

11.

Krishnan,

, Ranganathan,

, Kantak,

S. S.

, Dhaher,

Y. Y.

, & Rymer,

W. Z.

(2012). Active robotic training improves locomotor function in a stroke survivor. Journal of NeuroEngineering and Rehabilitation, 9(1), 57.

12.

Lang,

C.E.

, Macdonald,

J.R.

, Reisman,

D.S.

, Boyd,

L.A.

, Kimberley,

T.J.

, Schindler-Ivens,

S.M.

, Hornby,

T.G.

, Ross,

S.A.

, & Scheets,

P.L.

(2009). Observation of amounts of movement practice provided during stroke Rehabilitation. Archives of Physical Medicine and Rehabilitation, 90(10), 1692–1698.

13.

Lo,

A.C.

, Guarino,

P.D.

, Richards,

L.G.

, Haselkorn,

J.K.

, Wittenberg,

G.F.

, Federman,

D.G.

, Ringer,

R.J.

, Wagner,

T.H.

, Krebs,

H.I.

, Volpe,

B.T.

, Bever,

C.T.

, Bravata,

D.M.

, Duncan,

P.W.

, Corn,

B.H.

, Maffucci,

A.D.

, Nadeau,

S.E.

, Conroy,

S.S.

, Powell,

J.M.

, Huang,

G.D.

, & Peduzzi,

(2010). Robot-assisted therapy for long-term upper-limb impairment after stroke. New England Journal of Medicine, 362(19), 1772–1783.

14.

Marchal-Crespo,

, & Reinkensmeyer,

D.J.

(2009). Review of control strategies for robotic movement training after neurologic injury. Journal of NeuroEngineering and Rehabilitation, 6(1), 20.

15.

Mazzoleni,

, Sale,

, Tiboni,

, Franceschini,

, Carrozza,

M.C.

, & Posteraro,

(2013). Upper limb robot-assisted therapy in chronic and subacute stroke patients: a kinematic analysis. American Journal of Physical Medicine & Rehabilitation, 92(10), e26–37.

16.

Mehrholz,

, Pohl,

, Platz,

, Kugler,

, & Elsner,

(2018). Electromechanical and robot-assisted arm training for improving activities of daily living, arm function, and arm muscle strength after stroke Cochrane Database of Systematic Reviews Electromechanical and robot-assisted arm training for improving activi. September.

17.

Miyai,

, Suzuki,

, Hatakenaka,

, & Kubota,

(2006). Effect of body weight support on cortical activation during gait in patients with stroke. Experimental Brain Research, 169(1), 85–91.

18.

Nichols-Larsen,

D.S.

, Clark,

P.C.

, Zeringue,

, Greenspan,

, & Blanton,

(2005). Factors influencing stroke survivors’ quality of life during subacute recovery. Stroke, 36(7), 1480–1484.

19.

Reinkensmeyer,

D.J.

, Wolbrecht,

E.T.

, & Bobrow,

J.E.

(2007). A computational model of human-robot load sharing during robot-assisted arm movement training after stroke. Proceedings of the International Conference of the IEEE Engineering in Medicine and Biology, 4019-4023.

20.

Rodgers,

, Bosomworth,

, Krebs,

H.I.

, van Wijck,

, Howel,

, Wilson,

, Aird,

, Alvarado,

, Andole,

, Cohen,

D.L.

, Dawson,

, Fernandez-Garcia,

, Finch,

, Ford,

G.A.

, Francis,

, Hogg,

, Hughes,

, Price,

C.I.

, Ternent,

,... Shaw,

(2019). Robot assisted training for the upper limb after stroke (RATULS): a multicentre randomised controlled trial. The Lancet, 394(10192), 51–62.

21.

Sivan,

, O’Connor,

R.J.

, Makower,

, Levesley,

M.C.

, & Bhakta,

(2011). Systematic review of outcome measures used in the evaluation of robot-assisted upper limb exercise in stroke. Journal of Rehabilitation Medicine, 43(3), 181–189.

22.

Waddell,

K.J.

, Birkenmeier,

R.L.

, Moore,

J.L.

, Hornby,

T.G.

, & Lang,

C.E.

(2014). Feasibility of high-repetition, task-specific training for individuals with upper-extremity paresis. American Journal of Occupational Therapy, 68(4), 444–453.

23.

Washabaugh,

E.P.

, Guo,

, Chang,

C.-K.

, Remy,

C.D.

, & Krishnan,

(2019). A portable passive rehabilitation robot for upper-extremity functional resistance training. IEEE Transactions on Biomedical Engineering, 66(2), 496–508.

24.

Washabaugh,

E.P.

, & Krishnan,

(2018). A wearable resistive robot facilitates locomotor adaptations during gait. Restorative Neurology and Neuroscience, 36(2), 215–223.

25.

Washabaugh,

E.P.

, Treadway,

, Gillespie,

R.B.

, Remy,

C.D.

, & Krishnan,

(2018). Self-powered robots to reduce motor slacking during upper-extremity rehabilitation: a proof of concept study. Restorative Neurology and Neuroscience, 36(6), 693–708.