If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author Bulmaro Adolfo Valdés, MPE, Robotics for Rehabilitation Exercise and Assessment in Collaborative Healthcare Lab, Department of Mechanical Engineering, 6250 Applied Science Lane, The University of British Columbia, Vancouver, BC V6T 1Z4, Canada.
Robotics for Rehabilitation Exercise and Assessment in Collaborative Healthcare Lab, Department of Mechanical Engineering, The University of British Columbia, Vancouver, BC, Canada
Robotics for Rehabilitation Exercise and Assessment in Collaborative Healthcare Lab, Department of Mechanical Engineering, The University of British Columbia, Vancouver, BC, Canada
To investigate whether the compensatory trunk movements of stroke survivors observed during reaching tasks can be decreased by force and visual feedback, and to examine whether one of these feedback modalities is more efficacious than the other in reducing this compensatory tendency.
Design
Randomized crossover trial.
Setting
University research laboratory.
Participants
Community-dwelling older adults (N=15; 5 women; mean age, 64±11y) with hemiplegia from nontraumatic hemorrhagic or ischemic stroke (>3mo poststroke), recruited from stroke recovery groups, the research group's website, and the community.
Interventions
In a single session, participants received augmented feedback about their trunk compensation during a bimanual reaching task. Visual feedback (60 trials) was delivered through a computer monitor, and force feedback (60 trials) was delivered through 2 robotic devices.
Main Outcome Measures
Primary outcome measure included change in anterior trunk displacement measured by motion tracking camera. Secondary outcomes included trunk rotation, index of curvature (measure of straightness of hands' path toward target), root mean square error of hands' movement (differences between hand position on every iteration of the program), completion time for each trial, and posttest questionnaire to evaluate users' experience and system's usability.
Results
Both visual (−45.6% [45.8 SD] change from baseline, P=.004) and force (−41.1% [46.1 SD], P=.004) feedback were effective in reducing trunk compensation. Scores on secondary outcome measures did not improve with either feedback modality. Neither feedback condition was superior.
Conclusions
Visual and force feedback show promise as 2 modalities that could be used to decrease trunk compensation in stroke survivors during reaching tasks. It remains to be established which one of these 2 feedback modalities is more efficacious than the other as a cue to reduce compensatory trunk movement.
These augmented feedback strategies offer advantages when compared with trunk restraint: the person makes a conscious choice not to compensate, rather than relying on physical restraints that continuously limit body movement; it is less intrusive because there is no need to restrain the person to a chair; it can be used at home without direct supervision; the feedback intensity can be modified in real time from a remote location; and the active error thresholds and challenge of the task can be automatically adapted as the individual improves. To adopt augmented feedback in common rehabilitation practice, there must be sufficient evidence supporting the efficacy of these alternate feedback methods. In this study, we used 2 augmented feedback modalities (visual and force) to provide information to participants about their trunk compensation. The objectives of this study were (1) to investigate whether the compensatory trunk movement of stroke survivors can be decreased by force and visual feedback during reaching tasks; and (2) to investigate whether one of these feedback modalities is more effective in reducing compensatory trunk movement.
Methods
Participants
Fifteen participants were recruited (fig 1) from stroke recovery groups, the research group's website, and from the community. Table 1 provides a summary of the participants' demographics. A previous controlled trial
that investigated the reduction of stroke survivors' trunk compensation using trunk restraint provided the rationale for the chosen sample size. Participants provided written consent, and the study was approved by the clinical research ethics board.
Fig 1Recruitment and allocation (January to April 2016). The enrollment, allocation, and assignment of participants were conducted by the first author. The allocation sequence was stored on a digital file, and the participants were not aware of their allocation until after the familiarization with the system was completed and the baseline measurements were taken.
Baseline impairment and compensation assessments were administered by occupational therapists to determine the clinical characteristics of the participants (see table 1). The UE subsection of the Fugl-Meyer Assessment (FMA)
was used to assess level of participant compensation when reaching forward.
Experimental design and randomization
The trial used a crossover design; all participants experienced both treatments, and the order of the treatments was randomized. To reduce order effects, participants were randomly allocated (computerized pseudorandom number generatora) to start with visual or force feedback (see fig 1). Participants were first stratified according to FMA impairment scores (moderate to severe [<50] and mild [≥50]
) to ensure group balance, and then randomly allocated to the 2 treatment groups in blocks of 2. Included in the final analysis were 8 participants allocated to start with visual feedback, and 7 with force feedback. Figure 2 details the experimental procedure.
Fig 2Experimental design. Number of trials in parenthesis. Participants did not receive feedback in any of the posttrials. This was a low-risk study, with fatigue being the only possible harm. To reduce fatigue, participants received 1 minute of rest after every 15 trials, and were able to rest between targets if requested. An average of 17.3±6.8 minutes elapsed between the end of the first feedback condition and the start of the second one. Abbreviations: Condition A, visual feedback; Condition B, force feedback.
The integrated system (fig 3) consisted of 2 JACO2 (6DOF) robotic arms,b a Kinect 2 motion tracking camera,c and a personal computer. The system was controlled through a custom LabVIEWd program that displayed the reaching task on a monitor. Participants sat on a chair with at least 75% of their thighs resting on the seat, and a backrest and footrest adjusted to keep their hips and knees flexed at 90°.
Fig 3Experimental setup. Participants moved the robotic devices while completing the reaching task (displayed in the computer monitor). In addition, a motion tracking camera was placed in front of the participant to monitor trunk compensation.
Participants were instructed to move 2 virtual cursors (fig 4) representing each of their hands toward a target, and stay inside target bounds for 1 second. To move the cursors, participants performed symmetric bimanual reaching movements from their hips to their knees (without touching their thighs), while holding two robotic device handles (see fig 3). Moving the robots required minimal resistance because both robot arms were under admittance control
(robot sensed applied force and moved in the same direction). After every trial, participants returned to their initial calibrated position. If participants were unable to hold the robots' handles, they were provided with a wrist splint and a strap.
Fig 4Virtual reaching task with visual feedback active. Participants needed to move both cursors inside the target (2 horizontal lines, top of the figure) to complete 1 trial. When not receiving visual feedback, the cursors would be empty (white).
The movement of the cursors was only mapped to the anterior/posterior movement of participants' hands, and the robotic devices were capable of moving in 2 directions (up/down and forward/backward) (Video 1 in supplemental appendix S1, available online only at http://www.archives-pmr.org/). Participants were told that moving up/down would not affect the task, and that they should aim to move their hands at a constant height above, and close to, their thighs.
The distance to the virtual target (90% of hip-knee distance) was calibrated by asking participants to move their unaffected hand from their hips to their ipsilateral knee. Before the session started, participants were asked to push as hard as possible, with the robotic arms stationary, to ensure that the maximum torque that they could exert was above the maximum force feedback that they would receive (9.5Nm based on robots' torque limits). This torque was equivalent to the force required to hold a 1.23-kg object. Pilot studies had shown that this force is easily perceived by healthy participants. To ensure that participants with stroke could sense the force, all participants confirmed during familiarization trials that they could feel how the force changed as they compensated with their trunk.
The robotic arms provided force feedback when the Kinect motion tracking camera detected that the participant showed anterior trunk displacement during a reaching movement. The feedback adjusted the minimum torque required to move the robotic arms. This type of feedback was chosen because it provided a safety advantage; the robots would not move unless the participants actively moved them, whereas a purely resistive force acting in the opposite direction of motion could harm the participant if they released the robots' handles. Up to the first 30mm of compensation, participants did not receive feedback because they were considered to be within the normative threshold of healthy compensation.
After this threshold, the force feedback was proportional to the amount of trunk compensation (fig 5), and saturated at 50% of the average compensation each participant exhibited at baseline, minus the healthy compensation. The desired compensation was then set to 50% to promote achievable improvement in a short-term intervention. Our study involved only 1 training session; as a result, the desired compensation value was set to a static value. However, for interventions with multiple sessions, this value could be adjusted by researchers after every session to adapt to the current progress of their participants.
Fig 5Provided feedback calculation. The maximum force feedback was 9.5Nm, and the minimum was 1Nm. The maximum visual feedback was 100%, and the minimum was 0%. Abbreviations: F, force feedback; Max., maximum; Min., minimum; V, visual feedback.
The visual feedback operated using the same algorithm as the force feedback (see fig 5), and was represented as red ink filling up the virtual cursors, similar to a thermometer filling up, and proportional to the amount of trunk compensation (see fig 4). In this condition, participants also moved the cursors using the robotic arms, but the force feedback was turned off. This visual display was chosen because it did not add a new element to the screen (avoiding adding to the users' cognitive load), participants would already be familiar with this type of symbol, and it did not require detection of color change, which would be an issue for color-blind people.
Data analysis
All kinematic variables analyzed were measured during baseline and after visual and force trials, in which participants were not receiving feedback. The motion data were obtained from the Kinect and JACO arms at approximately 30Hz. The data were then resampled at a constant rate (25Hz), and low-pass filtered (6Hz
). If any of the Kinect's data points were inferred or not tracked, they were removed from the motion log. The Kinect's spine-shoulder and shoulder joints have been reported to have an average accuracy of approximately 10±10mm, with a high correlation (.99), when compared with a criterion standard motion capture system.
The primary outcome was trunk displacement (anterior displacement of the Kinect's spine-shoulder joint). Secondary outcomes are as follows. The first included trunk rotation: angle between the vector created from the left to the right shoulder joints, and the frontal plane (positive angles indicate counterclockwise rotations). The second included index of curvature: measure of the straightness of the hands' path toward the target in the y- and z- (superior/inferior and anterior/posterior) directions. The index was defined as the ratio of the hands' path and a straight line. A value of 1 would represent a perfectly straight path. The third was the root mean square error in y and z: measure of bimanual symmetry between the hands' movement. This error was computed as the difference between the hands' position at every iteration of the program, and the root mean square error of these values was calculated to obtain the final result. Smaller errors indicated more symmetric movements. The fourth was time: measured from the moment participants were presented with the reaching task to the end of the trial. The final measure was the posttest questionnaire: administered at the end of the study to investigate the experience of the participants and the usability of the system using the System Usability Scale.
To investigate whether there were any differences between visual and force feedback to reduce compensation, an analysis of covariance (ANCOVA) was used with a within-subject factor of treatment (visual or force), a between-subjects factor of group (start with visual or force), and the baseline measurements used as a covariate. To elucidate whether force and/or visual feedback reduced trunk compensation, the percentage gains (percent change from baseline to postmeasurements) were compared against a mean value of 0 using a 1-sample t test. When data violated parametric assumptions, the nonparametric sign test was used. For post hoc tests, the P values were adjusted using the Bonferroni-Holm correction.
When comparing visual against force feedback (table 2), for all outcome measures, all the main effects and interactions of the ANCOVA were not statistically significant (P>.05). The only exception was the left index of curvature where the interaction between treatment and baseline was significant (P=.001), which would invalidate the results from the ANCOVA's significant treatment effect (P=.002) for this measure. Therefore, for the outcome measures used in this study, there is no evidence that one feedback method is more effective.
Table 2Comparison between postforce and postvisual variables
Outcome Measure
Baseline
Postvisual
Postforce
Trunk displacement (mm)
119.2±71.7
69.8±73.1
68.7±64.6
Trunk rotation (deg)
−1.2±6.0
−2.2±7.2
−1.5±6.5
Time (s)
7.4±4.2
5.5±1.5
5.7±2.1
Index Curv. left yz
1.3±0.67
1.1±0.24
1.1±0.15
Index Curv. right yz
1.5±1.4
1.1±0.13
1.2±0.23
RMS z (mm)
22.4±11.4
31.1±29.0
29.7±23.9
RMS y (mm)
31.6±25.2
40.6±25.4
35.9±20.3
NOTE. Values are presented as mean ± SD. ANCOVA was used to compare postvisual versus postforce.
Abbreviations: Curv., curvature; RMS, root mean square.
When investigating if visual and force feedback reduced trunk compensation from baseline (table 3), a significant (P=.004) large effect (.99 and .89, respectively) was observed for both methods. Individual results are presented in table 4. For visual feedback, 8 of 15 participants reduced their compensation by >50%, 10 of 15 participants reduced their compensation by >30%, and 2 of 15 participants increased their compensation by <33%. For force feedback, 8 of 15 participants reduced their compensation by >50%, 8 of 15 participants reduced their compensation by >30%, and 3 of 15 participants increased their compensation by <30%. This evidence suggests that augmented visual and force feedback can reduce trunk compensation in hemiparetic stroke survivors. For all other measures, the differences were not statistically significant. Posttest questionnaire results are presented in supplemental appendix S1.
Table 3Comparison between percentage change from baseline to postmeasurements
NOTE. Values are median (first, third quartiles) or as otherwise indicated. The t test and sign test (values reporting median and quartiles) were used for percentage change comparisons.
Both visual and force feedback decreased trunk compensation exhibited by stroke survivors after a session of reaching trials with augmented feedback provided in these modalities. When comparing force with visual feedback to reduce trunk compensation, we did not find any significant differences. In addition, when asked if receiving visual or force feedback reduced how much they moved their trunk, most participants agreed (93.3% and 100%, respectively). This suggests that regardless of the modality of augmented feedback, participants use this information to correct their movement in a similar manner. However, studies with larger samples should be conducted to confirm this hypothesis. The question of which feedback medium is most effective for UE rehabilitation remains unanswered.
These augmented feedback modalities offer advantages for unsupervised, remote, or intensive rehabilitation because they do not require a therapist to physically restrain the individual or provide feedback in real time; the system used in this study was composed of commercially available products that could be integrated to provide rehabilitation outside of a research/rehabilitation setting. The lack of a physical constraint could provide additional benefits because clients could make a conscious choice about controlling their trunk movement,
which is something that a physical constraint could impede. With the physical guidance provided by the trunk restraints, the clients might not actively plan/program their trunk movements, which could inhibit important efferent and afferent information necessary for creating the internal models of the movement.
The augmented feedback used in this study has the potential to be provided at different frequencies during UE rehabilitation exercises, offering a variable schedule of reinforcement. Inversely, the continuous nature of the feedback provided by trunk restraints could be detrimental for motor learning; the guidance hypothesis states that practicing movements with constant feedback can make the participant dependent on the feedback, hindering independence.
However, for stroke survivors who show severe motor impairment with very limited trunk control, trunk restraint might be the only safe and viable option. As these individuals recover trunk control, and internal representations of movement are acquired, rehabilitation should move toward augmented feedback exercises, progressing to an eventual removal of feedback in a graded manner.
In our study, 6 participants with greater UE motor impairment (FMA score ≤38) struggled to complete the force feedback condition because of UE weakness. These participants' affected hands had to be strapped, taped, or supported with a wrist brace to hold onto the robots' handles or/and keep their wrist in a neutral position while they pushed through the force. Conversely, there were participants who found visual feedback less helpful because it was easier to ignore, did not add any resistance to the movement, or was harder to understand. In the posttest questionnaire, 46.7% of participants responded they would prefer to receive both feedback conditions, 26.7% would prefer to receive only visual, and 26.7% would prefer to receive only force. These observations, combined with the finding that there was not a statistical difference between using visual or force feedback, suggest there may not be an ideal feedback modality that works for every stroke survivor. We should instead use technology to provide feedback in an individualized manner, working to find the most suitable modality for an individual's impairment level, recovery stage, and learning style. Moreover, varying or combining the feedback medium could be most effective for rehabilitation. By varying feedback type throughout exercises, we could prevent clients from relying on a particular source of information to correct their movements. Varying feedback in a random schedule ensures novelty, which is important for retention and transfer of motor learning.
In addition, the setting in which rehabilitation occurs should be considered because visual feedback could be easier and more cost-effective to implement using devices that are already available in the home (ie, television, computer monitor), and force feedback may be more suitable in clinics or hospitals where larger, more costly devices can be acquired.
In this study we did not investigate whether a simple verbal instruction to avoid compensatory movement would effectively decrease compensation. To mitigate this limitation, we used the same number of repetitions (60) and a similar experimental procedure as a previous stroke rehabilitation controlled trial,
in which investigators compared a verbal instruction with a trunk restraint group on a unimanual physical reach-to-grasp task. The number of participants (N=14) and the samples were similar; however, our participants were on average older and more impaired (FMA scores). The previous study found that verbal instructions did not reduce compensation, whereas trunk restraint did. Our percent change values for visual (−41%) and force (−42%) feedback were on average superior to their trunk restraint values by 10% and 11%, respectively, and by 31% and 32% when compared with their verbal control condition. These results suggest that on average, augmented visual and force feedback in a short-term intervention could provide similar results to trunk restraint, and superior results to verbal instructions. Moreover, a study
investigating the use of visual feedback and operant conditioning in 5 video game rehabilitations sessions reduced relative compensation (trunk lateral lean) compared with no feedback. Further, a longer-term (12 sessions) bimanual/unimanual intervention
study investigating the use of auditory feedback versus trunk restraint found both methods improved scores on the RPS, FMA, and Wolf Motor Function Test. Our results, combined with these previous studies, suggest that augmented feedback could be used as a complement or substitute to trunk restraint.
Study limitations
The current pilot study investigated the effects of feedback in only a single session. Longitudinal studies should be conducted to explore the long-term effects of this intervention type. Because kinematic data alone are not sufficient to confirm the clinical utility of augmented feedback for rehabilitation, future studies should examine whether the changes in movement seen with these feedback modalities correlate with increased functional performance and independence with activities of daily living. Our results had large SDs because of the heterogeneity of the sample in terms of motor function, as shown by the baseline FMA and RPS scores. Studies with larger samples would enable researchers to stratify participants to various groups based on motor impairment. This approach would allow researchers to draw stronger conclusions about the effects of augmented feedback on stroke survivors with different motor/functional abilities. In our study, some participants had to use a wrist brace and/or strap to hold onto the robotic arm. Future studies should investigate alternate approaches to secure the hands while minimizing any potential effects to the participants' reaching performance. Finally, the force feedback that participants received was sensed through their upper limbs while holding the device's handles, which limits the generalization of these results to one sensing area of the body. It should be investigated if providing feedback cues directly to the trunk through a haptic or vibrotactile device could result in improved results to the ones presented in this work.
Conclusions
Both visual and force feedback appear to be effective candidates for reducing trunk compensation of stroke survivors. It remains to be established whether one of these feedback modalities is more efficacious. Using technology to provide real-time feedback that works best for each individual may be more effective than using one modality for all individuals who exhibit trunk compensation poststroke.
Suppliers
a.
Blocked randomisation list; Sealed Envelope.
b.
JACO2 (6DOF) robotic arms; Kinova Robotics.
c.
Kinect 2 motion tracking camera; Microsoft.
d.
LabVIEW; National Instruments.
Acknowledgments
We thank our colleagues Ashea Neil, Keith Lohse, Navid Shirzad, Tina Hung, Stephanie Glegg, Jonathan Marr, Laura Jaquez, Derek Schaper, and Renee Bernard for their assistance.
Usability testing of gaming and social media applications for stroke and cerebral palsy upper limb rehabilitation.
(In: 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). 2014 Aug 26-30; Chicago IL. New York, NY)2014: 3602-3605
Supported by the Peter Wall Solutions Initiative (grant no. 11-079), Vancouver, Canada; Consejo Nacional de Ciencia y Tecnología Mexico (grant no. 311462); and Canada Foundation for Innovation.