Abstract
Stroke is a leading cause of disability worldwide with improved treatments leading to higher survival rates and a greater demand for post-stroke rehabilitation. Technological rehabilitation solutions show promise to meeting this growing need for therapy, but for maximum impact they must be accessible and affordable. Here we investigate the potential for REVIVE, a low-cost virtual reality (VR) stroke rehabilitation system, to meet these requirements. The software we developed operates on the standalone Meta Quest line of VR Head Mounted Displays which are wireless, low cost (<$300 USD for the base 3S model), and require no external hardware. Hand tracking and voice recognition enable accessible engagement for users with a range of physical and cognitive abilities. An animated coach in the virtual environment guides users through gamified exercises simulating activities of daily living and functional movements. We tested the system with 60 younger adults without disabilities in a simulated stroke rehabilitation session. Users reported minimal visually induced motion sickness and more positive attitudes towards the VR technology after the experience. Additionally, we provide a normative dataset for several REVIVE tasks to serve as a healthy baseline for future clinical applications. Our findings suggest that this affordable consumer-grade system, suitable for home or clinical use, has considerable potential to improve access to post-stroke rehabilitation services.
Introduction
Worldwide, stroke ranks among the top causes of disability (1). Increasing survival rates over the past 30 years due to improved treatments have led to lower death rates but more years lived with disability (2), indicating a growing demand for post-stroke rehabilitation care. But patients typically receive less care than they need (3).
Access to care is a barrier to maximal recovery, with barriers that include cost, distance to treatment centers, and limited transportation (4). These accessibility barriers are particularly acute in rural settings and in developing nations. In Canada, rural patients were less likely to receive stroke unit care (36% versus 51%), to see a neurologist (26% versus 65%), or to see an occupational therapist (52% versus 72%) compared to urban patients (5). This mirrors treatment disparity globally with patients in Low- or Middle-Income Countries (LMICs) being less likely to receive standard care recommendations (6,7).
Increasing access to intensive post-stroke rehabilitation therapies is an important goal that could have a substantial impact on patient outcomes worldwide. Telehealth and other technology-enabled solutions can help overcome these barriers to treatment and have been shown to be as effective and cost-efficient as in-person stroke care (8), enabling an increase in therapeutic dosing and access (9). Public acceptance of telehealth solutions is high (9), including among adults 65 years and older who make up about 75% of stroke cases (10). Limitations persist as technology-enabled stroke rehabilitation options can place additional time and training burdens on clinicians (11), be costly (12) and/or require multi-component hardware setups (13,14).
Efforts to use low-cost VR headsets, such as the Meta Quest 2, to deliver stroke rehabilitation have demonstrated beneficial impacts on stroke recovery (15,16). Some evidence suggests that immersive VR confers greater benefit than conventional rehabilitation in measures of shoulder, wrist and hand function (17), and gross motor function (18). Dynamic adjustments to task parameters enable these systems to adapt to the different abilities of users, for example, shifting interactable objects closer to a user with limited range of motion (19,20). These approaches typically rely on changing the environment, whereas an alternative approach, motor amplification, amplifies the user’s movements to reduce their motor deficit in the virtual world. Motor amplification has been shown to reduce paretic limb non-use in stroke patients (21) and offers some benefits over environmental manipulation by maximizing access to users with a wide range of motor abilities since it can be applied to movement dynamics beyond motor extent (i.e. position in space), including the extent of hand grasp and wrist rotation. Attempts to implement amplification on low-cost headsets have been limited, with one attempt requiring the need for controllers and manual setting of parameters (22), reducing its ease of usability. Furthermore, feasibility and acceptance of VR rehabilitation technology has been underexplored. It is unclear if patients will be able to use these systems with limited oversight and if they will remain compliant with prescribed exercises. Users of a VR smoothie bar game reported high levels of engagement and satisfaction (23), but insight into other aspects of the patient experience are needed to inform the feasibility and usability of VR stroke rehabilitation, including levels of induced motion sickness and acceptance of the technology, particularly in patient populations that skew older.
To fill a need for more affordable and user-friendly stroke rehabilitation options, we developed the REVIVE virtual reality system (24) (Fig. 1). The REVIVE system (Rehabilitation and Engagement through Virtual Immersive Environments) uses a head-mounted display and an animated virtual coach to guide patients through exercises, reducing the need for constant clinician presence. Features include gamified tasks for high-dose motor exercise, hand tracking to eliminate the need for handheld controllers, and wireless operation. One noteworthy feature is a MotorAssist algorithm that amplifies user movements in VR, temporarily reducing post-stroke motor deficits to facilitate task completion and reduce frustration, thus leveraging the power of reinforcement learning (21,25). We advance previous motor amplification implementations by adding automated calibration operating without controllers on a low-cost headset. The REVIVE software runs on a commercially available headset, the Meta Quest 2/3S/3, which costs about the same as four physiotherapy sessions, enabling widespread deployment.
REVIVE system overview A. User interacting with virtual environment through a Meta Quest 2 head mounted display. B. User view within the virtual environment during the Bubble Pop Game. C. User view during the Pong Game. D. User view during the Animal Feeding Task.
Here we present the results of feasibility testing with 60 younger adults without disability, along with a normative baseline dataset that can be used to make clinical comparisons. The primary focus of the feasibility testing was to determine participant receptiveness to the technology as measured by a VR attitudes survey before and after the simulated rehabilitation session. Demonstrating the technology as accessible is critical to increase its acceptance by patients, particularly in individuals who may have less experience with emerging technologies including older adults (26). We also elaborate on the prevalence of visually induced motion sickness (24), which could be a barrier to the adoption of VR-based stroke rehabilitation. Our findings will inform the ongoing development of the REVIVE system and contribute to the ongoing discourse related to low-cost rehabilitation options for stroke, particularly in LMICs and rural areas with reduced access to care (6,7).
Methods
REVIVE System
We created a software program featuring 12 gamified tasks, with three tested in this study. These tasks were designed to either simulate daily activities, like pouring water into a glass, or to act as immersive games where the hand acts as a controller in such classic games as Pong or Snake. Each task targeted one of three movement categories: 1) Gross reaching; 2) Reach and grasp; or 3) Precision grasping.
In this study, the software was run on the Meta Quest 2 head mounted display, an immersive virtual reality system that can operate wirelessly (the software also runs similarly on the more recently released Meta Quest 3S and 3). Onboard cameras track the hands of the user, allowing the user to interact with objects in the virtual space without the need for handheld controllers or external camera setups. Real-world physics are simulated in the virtual environment using the Unity Game Engine. Some object interactions behave differently from their real-world analogue, such as grasping an object, and these incongruencies are demonstrated and explained in an initial tutorial for first-time users. We did not run the interaction tutorial in this study since the tasks completed in this study involved only gross reaching movements. At the beginning of each session, the user completes a guided calibration protocol which measures the range of motion (ROM) of each arm and hand. Any bilateral difference in ROM can be reduced by amplifying the movements of the deficient limb in the virtual world. For example, if a stroke survivor has a reduced range of shoulder flexion in one limb in the real world, the tracked movements can be amplified and displayed to match the unaffected limb’s range of motion. We apply a motor smoothing algorithm to reduce the jitteriness induced by amplified hand tracking noise.
Usability Study
Following the initial development of the REVIVE system (24), we conducted preliminary testing with healthy young adult participants to simulate a stroke rehabilitation session. This study was approved by the Acadia University Research Ethics Board (REB File #22-07). Sixty students from Acadia University receiving course credit for their involvement were recruited for the study. Participants provided demographic data, including gender identity [M/F/other], handedness [L/R/other], and age.
Study Setup
Participants donned a Meta Quest 2 VR headset connected to a desktop computer that ran the VR exercises. The experimenter monitored the virtual environment displayed to the participants and offered verbal guidance as necessary. The VR session lasted 10 to 15 minutes and included an initial calibration phase to assess participants’ range of motion. Surveys were administered to collect data on participants’ attitudes toward VR (pre- and post-session), prior technology use (pre-session), and motion sickness (post-session). Half of the participants experienced motor amplification of their hand movements, while the other half served as a control group with no amplification.
VR Task Overview
Participants began with a calibration task to establish their range of motion, after which motor amplification was applied for the experimental group. They then performed three tasks: a Bubble Pop game, a Pong game, and an Animal Feeding task. The order of the last two tasks was counterbalanced, with half of the participants in each group completing the Pong game first and the other half beginning with the Animal Feeding task. A description of each task is provided below:
Calibration Task
Participants reached for colored cubes positioned at different locations in their environment. The software recorded the range of motion for each hand and calculated the amplification factor for the experimental group based on the proportional difference between the two sides. This information was also used to set interaction boundaries for the subsequent tasks. Outcome measures included the volume of the ROM envelope and the amplification factor applied.
Bubble Pop Game
Participants popped virtual bubbles appearing in their field of view over a one-minute period. Hand positions were tracked at a frequency of 50 Hz to record movement trajectories. Outcome measures included the total number of bubbles popped, the overall distance traveled by the hands, and the 95th percentile hand velocity.
Pong Game
Participants controlled a virtual paddle by moving their hand forward or backward. The hand with the smaller range of motion was the controlled hand, with amplification applied in the experimental condition. A ball rolled across the table, and participants earned points by successfully deflecting the ball past the computer-controlled paddle. After a one-minute practice round, participants played a two-minute game. Outcome measures included the number of paddle-ball contacts, points scored, and opponent points scored.
Animal Feeding Task
Participants sat at a table in front of three cans of animal food arranged in a line from left to right. Each can was adorned with the image of an animal (dog, cat, bird) to indicate which animal’s food it contained. A virtual animal appeared at the far end of the table, and participants were required to push the matching can across a designated line away from the participant to “feed” the animal. The line’s distance was calibrated to the individual’s range of motion. Once the task was completed, the animal changed appearance and emitted an identifying sound (e.g. a bark for the dog), prompting the participant to push a new can across the line. Participants practiced for one minute before completing a two-minute session. The outcome measure was the number of correctly pushed cans.
Surveys
After providing written informed consent, participants completed surveys that collected demographic information and pre-intervention data on VR experience and attitudes. One question asked, “How often have you used virtual reality systems?”, with the options to select being Very frequently, Frequently, Occasionally, Rarely, Never. In the VR attitudes survey (Table I), participants responded to six statements using a Likert-scale [Strongly Agree (4), Agree (3), Neutral (2), Disagree (1), Strongly Disagree (0)]. The numerical equivalent scores reported for Statements 2 through 5 (Table I) were summed and their mean was used as an overall metric indicating attitudes towards VR technology. Scores for Statements 3 and 6 were reverse-coded.
Following the VR exercises, an abbreviated motion sickness symptom assessment was administered and the VR attitudes survey was readministered. The difference between the pre-survey and post-survey mean scores was used to indicate the overall change in VR attitude attributed to the VR experience. The motion sickness symptom assessment consisted of 5 of 9 items from the Virtual Reality Sickness Questionnaire (VRSQ) (27): General Discomfort, Fatigue, Eyestrain, Difficulty Focusing, Headache. The mean score across the five items was used to indicate visually induced motion sickness immediately following the VR experience.
Analysis Plan
Range of motion and motor performance scores on each of the three VR tasks were compared between groups using a two-sample t-test. For data presenting unequal variances, Welch’s version of the test was used. To determine if there was a change in VR attitude score across all participants following the VR experience, a Wilcoxon signed rank test was run (on the medians) with the alternative hypothesis that the true location is not equal to zero (median ≠ 0). A Wilcoxon signed rank test was also conducted on motion sickness level medians to determine if they exceeded minimal levels after the VR experience, with the alternative hypothesis that the true median is greater than 1. Wilcoxon rank sum tests were run for comparisons between groups for both VR attitude change and reported motion sickness. Medians (and thus the nonparametric Wilcoxon tests) were used with the non-normally distributed zero-inflated data that was expected. Analysis scripts in the R programming language along with all data are freely available at OSF: https://osf.io/45kad/?view_only=cc0f6a49e7dc49ae9af0392fb03262f0
Results
Sixty young adults without disability participated in the simulated VR rehabilitation exercises, either with or without motor amplification. We sought to generate a normative baseline of user performance, measure attitudes towards the technology and how they change with VR exposure, gather feedback on user experience including motion sickness symptoms, and identify any effect of amplification on these metrics.
47 participants identified as women, 10 as men and 3 as non-binary. Participants ranged in age from 18 to 39 years old (mean age = 19.7 years, SD = 3.22 years). Three self-reported as left-handed.
During the calibration task, a Welch two-sample t-test indicated that the range of motion (ROM) for the left hand (μ = 0.41m2, range: 0.13 – 1.17 m2) was not statistically different (t(117.97) = − 0.266, p = 0.79) from the right-hand ROM (μ = 0.42m2, range: 0.12 – 1.07 m2) (Fig. 2, Supplementary Fig. 1). No difference between hands was observed for the total distance moved during the calibration task and bubble pop game (Supplementary Fig. 2).
ROM calculated during calibration, broken down by hand and amplification status. Lines connect the left and right ROM for each individual participant. For amplified participants, the hand which had the lower extent in at least 2 of the 3 dimensions was selected to be amplified. For one right hand amplified participant (right panel), even with this criterion, the volume of the range of motion envelope was larger in the amplified side.
Total distance moved during the calibration phase and bubble pop task, broken down by hand and amplification status. White filled circles indicate group means.
Range of motion (ROM) by hand during calibration. See Supplementary Fig. 1 for a breakdown of ROM by hand and amplification status.
Based on an individual’s difference in ROM between the hands, an amplification factor was calculated to make these ROMs equal. The amplification factor calculated for participants with smaller left-hand ROMs (μ = 1.056, range: 1.003 – 1.201, n = 25) was slightly larger than those with smaller right hand ROMs (μ = 1.035, range: 1.002 – 1.164, n = 35), but this was not a statistically significant difference (t(38.19) = 1.77, p = 0.085, Welch’s t-test) (Fig. 3). After centering the amplification factors around zero (i.e. >0 indicates a right-hand amplification), a one-sample t-test indicated the amplification factors were not statistically different from zero (μ = −0.003, t(59) = −0.376, p = 0.71), indicating no hand side bias in detected ROM in this able-bodied participant sample.
Computed amplification factor for left and right hands across all participants.
With these small amplification factors, we observed no statistically significant effect of amplification on motor performance indicators across the three games tested (Figs. 4-6). In the Bubble Pop task, the amplified participant group popped a mean of 116.4 bubbles in one minute [SD = 66.3, range: 10 - 299] and the control group popped 102.5 bubbles [SD = 44.9, range: 13 - 235] (Fig. 4). A Welch’s t-test indicated this was not a statistically significant difference (t(50.99) = −0.95, p = 0.34).
Bubble pop results for participants without and with amplification.
In the Pong Game, the amplified participant group hit the ball a mean of 26.8 times in 2 minutes [SD = 3.00, range: 19 - 33] and the control group hit the ball 27.1 times [SD = 2.30, range: 23 - 32] (Fig. 5). A Welch’s t-test indicated this was not a statistically significant difference (t(54.32) = 0.39, p = 0.70).
Pong Game results for participants without and with amplification.
In the Animal Feeding Task, the amplified participant group pushed a mean of 38.1 cans in two minutes [SD = 21.1, range: 6 - 68] and the control group pushed 39.8 cans [SD = 18.8, range: 2 - 73] (Fig. 6). A Welch’s t-test indicated this was not a statistically significant difference (t(57.24) = 0.37, p = 0.74).
Animal Feeding results for participants without and with amplification.
Across all participants in both conditions, there was an average increase in attitude towards VR of 0.23 (SD = 0.54) (Fig. 7A), representing a small effect size (Cohen’s d = 0.44) that was statistically significant as indicated by a Wilcoxon signed rank test with continuity correction (V = 987, p < .001). The mean increases for both conditions were numerically equivalent at 0.23 per question change in VR attitude score (amplification group SD = 0.53, control group SD = 0.56) (Fig. 7B).
Attitude change towards VR following the VR experience. A. Mean per question score change from the pre- to post-survey for all participants. A Wilcoxon signed rank test with continuity correction showed this change to be statistically significant (V = 987, p < .001). B. Same data as in A broken down by intervention group, with and without motor amplification.
Immediately following the VR experience there was a mean motion sickness score of 0.57 (SD = 0.44, median = 0.4) across all participants (Fig. 8A). Three participants did not answer all questions on the motion sickness survey and their data have been removed from the analysis. A Wilcoxon signed rank test with continuity correction was run with the alternative hypothesis of u>1 given that scores above 1 would indicate more than a minor effect. This test did not show a statistically significant effect on motion sickness (V = 95.5, p = 1). A Wilcoxon rank sum test with continuity correction indicated that the difference in motion sickness scores between amplified (mean = 0.64, median = 0.6) and non-amplified (mean = 0.50, median = 0.4) participants was not statistically different (W = 326, p = .20) (Fig. 8B).
Motion sickness immediately following the VR experience. A. Mean score per statement for all participants. A Wilcoxon signed rank test with continuity correction did not show a statistically significant effect (u > 1; V = 95.5, p = 1). B. Same data as in A broken down by intervention group, with and without motor amplification [adapted from (24)].
Discussion
We tested a virtual reality stroke rehabilitation platform in healthy younger adults to assess the feasibility of its use and to generate a pool of normative motor performance data. These data, alongside forthcoming data from healthy older adults, will enable quantitative motor recovery comparisons in stroke patients.
Importantly, we observed no statistically significant differences in motor performance between the amplification and no amplification groups. Despite the small amplification factors in the healthy participants (90% of participants had amplification factors less than 10%), the amplification still presents a visual perturbation. We have verified that this manipulation does not inadvertently affect baseline motor performance in healthy younger adults.
Participant attitudes towards VR technology improved following the VR experience, with similar positive shifts observed in both the amplification and no amplification groups. These results suggest the feasibility for more widespread clinical adoption of such VR exercises.
Another important finding is that minimal motion sickness was reported, though slightly higher motion sickness levels were observed in the motor amplification group. This trend could foretell potential issues when greater amplification factors are used, such as in clinical populations. Testing with older healthy adults and stroke patients with increased amplification is necessary to explore this further.
As an initial feasibility assessment of this new technology, our study has several limitations that should be addressed in future work. The focus here on a healthy younger adult population, while necessary, restricts the generalizability of our findings to inform the system’s future clinical deployment. A forthcoming study testing healthy older adults as well as ongoing work with stroke patients will address this limitation. Additionally, the VR exposure in this study was limited to three tasks in a single session; longer sessions or multiple sessions could induce more motion sickness symptoms. However, motion sickness symptoms have been shown to occur quite quickly, with one study reporting an average nausea onset time of 62 seconds (28). Nevertheless, ongoing comprehensive testing is needed to assess the system’s feasibility when implemented across multiple sessions in clinical contexts.
Overall, the VR system presented here shows strong potential for clinical deployment as a stroke rehabilitation solution. Importantly, the use of a low-cost consumer headset could make this emerging treatment paradigm accessible to more patients than previous VR stroke rehabilitation systems. Before clinical deployment, next steps include a forthcoming feasibility study with healthy older adults and clinical studies with stroke survivors.
Funding details
This work was supported by ResearchNS under the New Health Investigator Grant program.
Disclosure statement
The authors report there are no competing interests to declare.
Footnotes
Figures improved, text edits throughout to improve clarity
References
Subject Area
Reviews and Context
0
Comment
0
TRIP Peer Reviews
0
Community Reviews
0
Automated Services
0
Blogs/Media
Author Videos