This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Serious Games, is properly cited. The complete bibliographic information, a link to the original publication on http://games.jmir.org, as well as this copyright and license information must be included.
Most stroke survivors continue to experience motor impairments even after hospital discharge. Virtual reality-based techniques have shown potential for rehabilitative training of these motor impairments. Here we assess the impact of at-home VR-based motor training on functional motor recovery, corticospinal excitability and cortical reorganization.
The aim of this study was to identify the effects of home-based VR-based motor rehabilitation on (1) cortical reorganization, (2) corticospinal tract, and (3) functional recovery after stroke in comparison to home-based occupational therapy.
We conducted a parallel-group, controlled trial to compare the effectiveness of domiciliary VR-based therapy with occupational therapy in inducing motor recovery of the upper extremities. A total of 35 participants with chronic stroke underwent 3 weeks of home-based treatment. A group of subjects was trained using a VR-based system for motor rehabilitation, while the control group followed a conventional therapy. Motor function was evaluated at baseline, after the intervention, and at 12-weeks follow-up. In a subgroup of subjects, we used Navigated Brain Stimulation (NBS) procedures to measure the effect of the interventions on corticospinal excitability and cortical reorganization.
Results from the system’s recordings and clinical evaluation showed significantly greater functional recovery for the experimental group when compared with the control group (1.53, SD 2.4 in Chedoke Arm and Hand Activity Inventory). However, functional improvements did not reach clinical significance. After the therapy, physiological measures obtained from a subgroup of subjects revealed an increased corticospinal excitability for distal muscles driven by the pathological hemisphere, that is, abductor pollicis brevis. We also observed a displacement of the centroid of the cortical map for each tested muscle in the damaged hemisphere, which strongly correlated with improvements in clinical scales.
These findings suggest that, in chronic stages, remote delivery of customized VR-based motor training promotes functional gains that are accompanied by neuroplastic changes.
International Standard Randomized Controlled Trial Number NCT02699398 (Archived by ClinicalTrials.gov at https://clinicaltrials.gov/ct2/show/NCT02699398?term=NCT02699398&rank=1)
After initial hospitalization, many stroke patients return home relatively soon despite still suffering from impairments that require continuous rehabilitation [
One of the latest approaches in rehabilitation science is based on the use of robotics and virtual reality (VR), which allow remote delivery of customized treatment by combining dedicated interface devices with automatized training scenarios [
We conducted a parallel-group, controlled trial in order to compare the effectiveness of domiciliary VR-based therapy versus domiciliary occupational therapy (OT) in inducing functional recovery and cortical reorganization in chronic stroke patients.
Participants were first approached by an occupational therapist from the rehabilitation units of Hospital Esperanza and Hospital Vall d’Hebron from Barcelona to determine their interest in participating in a research project. Recruited participants met the following inclusion criteria: (1) mild-to-moderate upper-limbs hemiparesis (Proximal MRC>2) secondary to a first-ever stroke (>12 months post-stroke), (2) age between 45 and 85 years old, (3) absence of any major cognitive impairment (Mini-Mental State Evaluation, MMSE>22), and (4) previous experience with RGS in the clinic. The ethics committee of clinical research of the Parc de Salut Mar and Vall d’Hebron Research Institute approved the experimental guidelines. Thirty-nine participants at the chronic stage post-stroke were recruited for the study by two occupational therapists, between October 2011 and January 2012, and were assigned to a RGS (n=20) or a control group (n=19) using stratified permuted block randomization methods for balancing the participants’ demographics and clinical scores at baseline (
The RGS integrates a paradigm of goal-directed action execution and motor imagery [
Experimental setup and protocol: (A) Movements of the user’s upper limbs are captured and mapped onto an avatar displayed on a screen in first person perspective so that the user sees the movements of the virtual upper extremities. A pair of data gloves equipped with bend sensors captures finger flexion. (B) The Spheroids is divided into three subtasks: hit, grasp, and place. A white separator line divides the workspace in a paretic and non-paretic zone only allowing for ipsilateral movements.(C) The experimental protocol. Evaluation periods (Eval.) indicate clinical evaluations using standard clinical scales and Navigated Brain Stimulation procedures (NBS). These evaluations took place before the first session (W0), after the last session of the treatment (day 15, W3), and at follow-up (week 12, W12).
Designing automated evaluation tools to be used at-home in a non-supervised setup could provide objective and frequent measurements of recovery, offering valuable information to clinicians and primary users, and driving autonomous rehabilitation technologies. We, therefore, developed the Automated Evaluation of Motor Function (AEMF), a VR-based evaluation scenario for the assessment of upper-limb motor function that was designed to operate under non-supervised conditions.
In order to assess proximal and distal motor function, the AEMF scenario is divided into two separated tasks. In task 1, participants were asked to perform planar wiping movements with their arms to clear a virtual surface covered with small cubes. In task 2, participants were instructed to squeeze a virtual object by flexing and extending their fingers. In order to guarantee that the AEMF tasks were correctly understood, each of these was first performed using the non-paretic limb and then with the paretic limb. Participants did not receive any explicit feedback (ie, knowledge of results) about their overall performance. During task execution, we collected data of hand position and joint rotation (fingers, elbows, and shoulders) to compute three main performance descriptors: the horizontal planar area covered, finger flexion, and extension.
In order to test the effectiveness of VR in the domiciliary context, each participant received daily home-based upper-limb rehabilitation during 5 weekly days, for 3 consecutive weeks. The RGS group followed a home-based training paradigm based on the Spheroids scenario (
All participants’ motor function was evaluated at day 1, day 15 of the rehabilitation program, and week 12 follow-up (
Both during the training and evaluation sessions, we captured the user’s movements and mapped them onto a biomechanical model of the upper limbs. Specifically, virtual movements were controlled by the angles of the users’ joints measured by a motion capture device at 30Hz (Kinect, Microsoft, USA). The range of finger flexion was captured by a pair of data gloves (DGTech Engineering Solutions, Bazzano, Italy) equipped with bend sensors, measures range from 0 to 1, indicating maximum extension and maximum flexion respectively.
Navigated Brain Stimulation (NBS) procedures [
Navigated Brain Stimulation (NBS) procedure. Bottom right: axial and coronal view of a magnetic resonance imaging (MRI) scan at the level of the stroke for one of the participants in the experimental group showing a partial anterior circulation infarct due to an embolism. Bottom right: Example of NBS mapped cortical motor representations; colored areas indicate the targeted cortical sites.
For statistical analysis, data were tested for normality using the Kolmogorov-Smirnov test. To identify significant time effects on clinical scores we performed a Friedman test. Next, we conducted a post-hoc analysis using 2-tailed Mann-Whitney
In order to validate the RGS Adaptive Difficulty Controller, automatic performance ratios and difficulty parameters assigned by RGS to the paretic and non-paretic limb were compared (Wilcoxon signed-rank test). Next, to explicitly study progress in performance, we averaged values for each difficulty parameter per session and performed a within-subjects time-series analysis of the means (Friedman test).
Data of hand position and joint rotation collected during performance in AEMF were filtered using a second order Butter-worth low- pass filter (cut-off at 6 Hz) reducing noise. In order to assess the participant’s motor function within AEMF, we calculated three performance descriptors for each extremity: (1) the work area was defined as the dorsal surface area of the movement space, while (2) finger flexion, and (3) extension were defined as the maximal and minimal metacarpal angles respectively, averaged across all fingers.
We tested AEMF sensitivity by examining between-limbs differences in descriptor values (ie, covered area, finger flexion and finger extension) for each subject (Wilcoxon signed-rank test). Next, in order to explore AEMF test-retest stability and sensitivity to capture improvement, we analyzed changes in descriptor values across sessions (Friedman test). In addition, we studied the relation between standardized clinical scores and AEMF measurements of motor function by computing a Spearman correlation coefficient for each descriptor and clinical scale at the corresponding evaluation period.
Finally, we compared the Stimulation Efficacy (SE) and the centroid location of the cortical motor areas representing APB and ECR in M1, for the pathological and non-pathological hemispheres (Wilcoxon sign-sum test). In order to extract training effects, we performed a within-subject analysis of the Stimulation Efficacy and the centroid location of the cortical maps in M1 before and after treatment (Wilcoxon sign-sum test). We used a Spearman test to study the correlations between NBS outcome measures and improvements in clinical scales.
Two-sided significance level for all statistical tests was defined as alpha=0.05.Data processing and statistical analysis were performed using Matlab 2013a (MathWorks, Inc.). Due to limited statistical power, we did not correct for multiple comparisons.
Participants’ demographics and scores from clinical scales at baseline.
Demographics | RGS (n=17) | Control (n=18) | |
Gender (female), n (%) | 9 (53) | 12 (67) | .59a |
Age, mean (SD) | 65.05 (10.33) | 61.75 (12.94) | .44b |
Affected side (left), n (%) | 11 (65) | 9 (50) | .58a |
Type (hemorrhagic), n (%) | 6 (33) | 6 (33) | .81a |
Oxford class (LACc/PACd/TACe) |
4/3/4 | 6/2/4 | .65a |
Days after stroke, mean (SD) | 1073.43 (767.74) | 798.06 (421.80) | .64b |
MMSE [ |
28.24 (2.33) | 28.22 (2.34) | .08b |
Hamilton [ |
3.71 (3.35) | 4.56 (3.24) | .40b |
Grip force, mean (SD) | 6.15 (5.04) | 5.94 (5.85) | .69b |
MRCf proximal [ |
3.47 (0.51) | 3.39 (0.61) | .76b |
MRC distal [ |
2.82 (1.19) | 3.17 (0.99) | .44b |
FMA [ |
42.94 (14.37) | 43.44 (13.48) | .89b |
CAHAIg [ |
52.82 (23.10) | 53.50 (22.51) | .95b |
Barthel [ |
89.53 (9.43) | 84.72 (14.19) | .48b |
Ashworth proximal [ |
1.24 (1.25) | 1.22 (1.31) | .97b |
Ashworth distal [ |
1.47 (1.51) | 1.00 (1.41) | .42b |
VASh shoulder [ |
1.59 (2.76) | 2.61 (2.64) | .13b |
aChi-square test.
bWilcoxon rank-sum test.
cLAC: Lacunar stroke.
dPAC: Partial anterior circulation stroke.
eTAC: Total anterior circulation stroke.
fMRC: Medical Research Council.
gCAHAI: Chedoke Arm and Hand Activity Inventory (version CAHAI-13).
hVAS: Visual Analog Scale.
In order to assess the impact of the RGS treatment, we conducted a repeated measures analysis of the functional recovery captured through standardized clinical scales. Analysis of participants’ demographics revealed no significant differences between groups at baseline (
Participants in the RGS group completed a variable total number of Hit (37.1, SD 18.4), Grasp (35.1, SD 17.0) and Place (34.2, SD 16.8) subtasks along the 3 weeks of treatment. All patients participating in the study were able to put the gloves on with assistance, and autonomously set-up and use the system until finishing the game. In order to assess whether the adaptive difficulty controller effectively provided customized training intensities that matched the participants’ capabilities, we explored inter-limb differences in mean performance ratios during training. Differences in performance showed a trend toward significance in Grasp and Place subtasks (
In order to study the RGS AEMF sensitivity, we compared measurements for the paretic and non-paretic limb. In addition, we explored the test-retest stability of these parameters. We observed that estimates of working area and maximal finger extension performed by the paretic limb in AEMF were significantly lower when compared to the non-paretic limb (
A: AEMF captures an improvement in finger flexion during treatment. Averaged movement profile of fingers excursion performed by one subject during one of the sessions. Units of finger flexion are expressed as a ratio of complete flexion. B: Mean changes in maximal finger flexion for all subjects in the RGS group across the three weeks of intervention, for both non-paretic (NPL) and paretic limbs (PL).
Effects of RGS treatment versus control on clinical scales within and between groups for the post treatment assessment at week 3 and the long-term follow up at week 12.
Assessment | RGS (n=17) |
Control (n=18) |
Between Groups | Effect size | ||||||
Improvement, mean (SD) | Improvement, mean (SD) | Cohen |
||||||||
End (Week 3) | ||||||||||
UE-FMa | 0.35 (1.62) | .43 | 1.22 (3.84) | .15 | .33 | −0.30 | ||||
CAHAIb | 1.53 (2.4) | .01 | −0.67 (6.01) | .90 | .05 | 0.48 | ||||
Barthel | 0.00 (1.87) | >.99 | 1.00 (2.87) | .25 | .44 | −0.41 | ||||
MRCpc | 0.06 (0.24) | >.99 | 0.11 (0.32) | .50 | .61 | −0.17 | ||||
MRCdd | 0.06 (0.43) | >.99 | 0.11 (0.47) | .63 | .74 | −0.12 | ||||
Aspe | 0.00 (0.35) | >.99 | 0.06 (0.24) | >.99 | .32 | 0.40 | ||||
Asdf | 0.12 (0.33) | .50 | 0.00 (0.34) | >.99 | .32 | 0.36 | ||||
Grip force | 0.41 (1.78) | .89 | 0.38 (2.65) | .47 | .57 | 0.01 | ||||
Hamilton | 0.88 (2.45) | .16 | 0.67 (1.57) | .13 | .66 | 0.10 | ||||
VAS-Sg | 0.41 (1.81) | .05 | −0.28 (1.90) | .69 | .63 | 0.37 | ||||
Follow-up (Week 12) | ||||||||||
UE-FM | −0.18 (3.50) | .82 | 1.39 (3.63) | .11 | .21 | 0.34 | ||||
CAHAI | −0.06 (6.51) | .74 | 0.44 (5.46) | .67 | .61 | −0.08 | ||||
Barthel | −3.30 (8.09) | .29 | −0.11 (3.98) | .92 | .74 | −0.50 | ||||
MRCp | −0.12 (0.78) | >.99 | 0.28 (0.46) | .06 | .06 | −0.62 | ||||
MRCd | 0.29 (0.77) | .25 | 0.17 (0.62) | 45 | .98 | −0.17 | ||||
Asp | 0.06 (0.65) | >.99 | 0.00 (0.34) | >.99 | >.99 | −0.12 | ||||
Asdf | 0.29 (0.59) | .13 | 0.00 (0.00) | >.99 | .03 | 0.70 | ||||
Grip force | 0.21 (1.45) | .73 | 0.23 (3.02) | .92 | .93 | −0.01 | ||||
Hamilton | 0.35 (2.34) | .70 | 1.11 (3.53) | .42 | .93 | −0.25 | ||||
VAS-S | 0.12 (2.06) | .92 | 0.78 (3.08) | .38 | .27 | −0.25 | ||||
aUE-FM: The upper extremity Fugl-Meyer Assessment.
bCAHAI: Chedoke Arm and Hand Activity Inventory (version CAHAI-13).
cMRCp: Medical Research Council for proximal muscles.
dMRCd: Medical Research Council for distal muscles.
eAsp: Ashworth Scale for proximal muscles.
fAsd: Ashworth Scale for distal muscles.
gVAS-S: Visual Analog Scale for Shoulder Pain.
In order to detect training-induced changes in the corticospinal system, we first characterized cortical regions in the primary motor area of the pathological and non-pathological hemispheres representing abductor pollicis brevis (APB) and extensor-carpi radialis (ECR) muscles. At baseline, the Stimulation Efficacy (SE) was significantly higher for the non-pathological hemisphere when compared to the pathological one (
SE increased significantly within subject after treatment in the pathological hemisphere (3.6, SD 8.60;
We observed a centroid displacement in the pathological hemisphere, which occurred after treatment both for the APB and the ECR muscle (
Effects of domiciliary rehabilitation therapy on corticospinal efficacy. (A) Change in mean Stimulation Efficacy for extensor-carpi radialis (ECR) in the damaged hemisphere (pathological) and the intact hemisphere (non-pathological). (B) Change in mean Stimulation Efficacy for abductor pollicis brevis (APB). (C) Centroid displacements after therapy along anterioposterior and mediolateral axis. (D) Correlation of absolute centroid displacements after therapy with improvement in CAHAI score after therapy.
We have studied the effectiveness of the RGS VR-based system for home-based motor rehabilitation of the upper extremities after stroke by conducting a controlled, longitudinal clinical trial assessing both functional and structural impact and comparing it to an OT task. We have shown that, at the chronic stage post-stroke, the remote delivery of customized self-managed motor training in VR environments may effectively induce motor gains and neuroplastic changes. Comparisons between groups suggest a superiority of VR compared with OT in domiciliary setups, however, this difference does not reach clinical impact. Our results highlight the potential of automated rehabilitation technologies for domiciliary neurorehabilitation, which so far has been an issue of some contention [
First, we validated the RGS Adaptive Difficulty Controller, which automatically provides for a limb specific customization of practice difficulty and intensity, and a progress-monitoring tool. We observed lower success rates during the execution of those subtasks involving distal movements (ie, Grasp and Place). Lateralized customization of task difficulty allowed for the maintenance of optimal performance levels for each limb across sessions. Within-subject analysis of the evolution of the difficulty parameters assigned during training revealed paretic limb specific functional improvements during a reaching and grasping task. These observations may indicate functional gains of distal function (ie, increased control in fingers flexion and extension). Data collected by the Automated Evaluation of Motor Function further confirmed these findings, revealing significant improvements for the paretic limb, during week 2 and 3, in finger flexion. Interestingly, we also found an improvement in range of movement both for the paretic and non-paretic limb, probably indicating a generalization of new cognitive and compensatory strategies. Notice that subjects included in this study were in the chronic phase of stroke (mean time post stroke 65.05 months, SD 10.3), a period in which motor improvements are supposed to have plateaued and limited non-compensatory functional gains can still be induced through further physical or OT [
Results from the NBS protocol supported these findings by displaying an enhanced corticospinal excitability after treatment only for the more distal muscle (ABP) associated with hand function. In addition, we observed centroid displacements of the cortical map for both the ABP and the ECR. This confirms earlier reports that enhanced corticospinal excitability and cortical map centroid displacements strongly correlate with functional gains detected by standardized clinical scales, such as Fugl-Meyer, CAHAI, and Barthel scales [
Taking a global perspective on these results, we observe that task difficulty descriptors, AEMF measurements, and NBS, converged, suggesting that distal functional improvements were induced through RGS based training and were significantly larger for those participants in the RGS group when compared with the control group. The reason why we may not have observed improvements in proximal muscle groups and other clinical scales may be related to the stringent inclusion criteria of the study, which excluded all subjects showing severe hemiparesis at baseline (Proximal Medical Research Council, MRC>2). It is widely known that the corticospinal system is organized following a proximal to distal gradient to the cervical spinal cord, where motoneurons of the distal muscle groups receive most input projections [
In this randomized controlled study, we explored the effects of a VR-based system for domiciliary rehabilitation on functional recovery and cortical reorganization. Our results suggest that at-home VR-based rehabilitation promotes functional motor gains, enhances corticospinal excitability, and induces cortical reorganization at the chronic stage post- stroke. The observation of strong correlations between increased motor evoked potentials after treatment and functional gains in CAHAI suggests that exposure to VR-based goal-oriented motor training may have enhanced the organization of corticospinal pathways, facilitating distal motor control. The displacement of the centroid of cortical maps after training may also indicate related cortical reorganization at the chronic stage post-stroke supporting the idea that recovery can be induced at any stage post stroke albeit to varying degrees.
CONSORT eHealth form.
automated evaluation of motor function
abductor pollicis brevis
Ashworth scale for distal upper limb
Ashworth scale for proximal upper limb
barthel index
chedoke arm and hand activity inventory
extensor-carpi radialis
mini-mental state evaluation
medical research council scale
navigated brain stimulation
occupational therapy
rehabilitation gaming system
simulation efficacy
the upper extremity Fugl-Meyer assessment
visual analog scale
virtual reality
We would like to thank all subjects who participated in this study. We also would like to gratefully acknowledge Estefanía Montiel for her assistance in recruiting and evaluating the participants. This work was supported by the MINECO “Retos Investigacion tos Investigacion I + D + I” Plan Nacional project SANAR (Gobierno de España), and the European Research Council under grant agreement 341196 CDAC and FP7-ICT- 270212 project eSMC.
PV is involved in the spin-off company Eodyne Systems SL, which has the goal to achieve a large-scale distribution of science based rehabilitation technologies.