Background
Chronic upper extremity impairment is all too common among the more than 7 million stroke survivors in the U.S. [
1]. These impairments have disabling effects on all facets of life, including self-care, employment, and leisure activities. Repetitive practice of movement, such as arm movement, is thought to improve outcomes for stroke survivors [
2‐
4], but access to the clinic for therapy is often limited by geography or lack of transportation. While almost 50 million Americans live in rural areas, 90% of physical and occupational therapists live in major urban areas [
5]. Per capita ratios of therapists to overall population are 50% larger in urban as compared to rural regions of the country [
6]. Rates of stroke in these rural areas, however, exceed those of major urban areas [
7‐
9]. Thus, a large number of stroke survivors have limited access to skilled treatment. Data from 21 states found that only 30% of stroke survivors received outpatient rehabilitation, a much lower percentage than that recommended by clinical practice guidelines [
10]. Declines seen following discharge from inpatient rehabilitation are undoubtedly exacerbated by limited access to clinical therapy [
11].
Disparity in quality of care has been recognized in the acute treatment of stroke for a number of years. This situation has led to the development of telemedicine to extend expert care to individuals during the initial hours and days following the stroke, advance site-independent treatment, and create models of care in rural areas [
12‐
14]. Therapy options after this acute period, however, generally remain limited for stroke survivors in rural areas. Akin to the telemedicine efforts, telerehabilitation treatments have been proposed. However, telerehabilitation interactions are typically limited to off-line monitoring by the therapist [
8,
9,
15], phone calls between a therapist and client [
16,
17], or videoconferencing [
18‐
20]. While systems allowing more direct interaction have been proposed, the hardware cost and complexity limit applicability for home-based therapy [
21‐
23]. Hence, the therapist is relegated to the role of observer and the intimacy of a clinical therapy session is lost. Therapy options are substantially restricted, as is the available feedback.
Recently, multiple investigators have been exploring means of improving home-based therapy through the development of systems or serious games which permit multiple, simultaneous users [
24‐
30]. These efforts have proposed the inclusion of multiple users as a means to overcome resistance to home-based therapy that may result due to isolation or lack of engagement. Indeed, studies have observed a preference for multi-user vs, single-user therapy when utilizing these systems [
26,
29]. However, these systems have largely been limited to control of a one-dimensional or two-dimensional space and both users remain in the same physical location (e.g., side by side). One team of researchers did develop a framework for supporting distant users (such as a therapist in the hospital and a stroke survivor in their home), but game control was limited to one or two dimensions [
31,
32].
Here, we describe the development of a fully three-dimensional (3D) virtual reality environment (VRE) for home-based therapy in which multiple, remote users can interact in real time. In this Virtual Environment for Rehabilitative Gaming Exercises (VERGE) system [
33], movement of the user is mapped to corresponding movement of an avatar to foster a sense of presence in and engagement with the VRE. The 3D environment encompasses aspects of clinical therapy, such as transport of objects or movement of the hand into specified regions of the upper extremity workspace. Although the importance of 3D movements in VR environments is a topic of debate [
34,
35], movements tested in environments with lesser degrees-of-freedom (DOF) are often very limited and dictated by a one DOF robot. These movements differ substantially from the types of movements normally seen in 3D reaching movements [
4,
36]. The network architecture of the system allows users to be located remotely from each other, such as a stroke survivor in their home, a therapist in a clinic, or a stroke survivor’s friend or relative living in another city or state. The virtual nature of the environment allows even very limited movements in the physical world to have successful functional outcomes in the virtual world, thereby offering a sense of accomplishment and motivation for successive attempts. Additionally, task difficulty can easily be modified in order to maintain the proper level of challenge, which is important for motor learning in general [
37] and rehabilitation in particular [
38].
We developed and performed preliminary testing of the VERGE system to gauge user response in comparison to two other therapy modalities that could be used for home therapy: an existing virtual reality system based on the Alice in Wonderland story (AWVR) [
39] and a home exercise program (HEP). Fifteen stroke survivors completed three, one-hour therapy sessions per week with each of the three therapy modalities (9 sessions total). We hypothesized that the use of the VERGE system would not decrease the amount of arm movement promoted, in comparison with the AWVR and HEP modalities. We further expected that users’ self-described engagement would be greatest for the VERGE system due to the presence of a partner.
Discussion
VERGE implementation
We developed a 3D, networked VR system allowing users, physically remote from each other, to interact within a virtual environment. Each user controls an avatar in real time by movement of corresponding body segments. These avatars can manipulate virtual objects located within the environment; multiple avatars can even manipulate the same object, such as a ball hit back and forth. Each user needs only have a computer, wireless mouse, and a Kinect™ device. No special software is required for the user, only an executable version of our code and the Kinect SDK.
This VERGE system was successfully tested by 15 stroke survivors with chronic hemiparesis in the upper extremity. User response was generally positive, with 85% of the participants expressing satisfaction with the utility of the therapy and 93% indicating satisfaction with the amount of arm movement induced. Indeed, participants moved their hands an average of 350 m (after subtracting shoulder translation) during each session. This far exceeds the amount of hand displacement produced by the 54 movements observed during a typical occupational therapy session [
50]. In accordance with previous multi-user training studies [
26,
29,
30], the vast majority (14 of 15) participants indicated that they liked having a partner for therapy, despite not being in visual contact with this person.
Comparison with other potential training modalities
Overall, all three therapy options encouraged considerable movement of the hand in space. Non-inferiority testing confirmed that use of the VERGE system did not result in significantly less displacement of the hand than that recorded using the more established AWVR or HEP modalities. As relatively few studies have quantified arm movement during therapy outside of a robotic device, these values provide an important target for therapy. Importantly, all three training modalities encouraged movement away from the body. Arm movements to areas of the workspace which require elbow extension can be challenging for stroke survivors, especially when the arm is unsupported [
51,
52], as was the case for VERGE, AWVR, and the majority of HEP. The differences observed in the amount of arm movement between modalities could be a result of confounding factors in exercise design. Specifically, exercises in VERGE were designed to include movements out of synergy and in a large free 3-D space. Although the importance of 3D movements in therapy is a topic of debate [
34,
35], many tasks require non-planar movements. VERGE allows practice of such task-based motions. AWVR also included movements out of synergy but the workspace was much more limited in size. HEP included many exercises with proximal arm stabilization, these movements were simpler in that they did not require multiple joint coordination or trunk stabilization.
During the training sessions, participants spent the most time with their arms extended at least 70% of full range. Participants also spent a considerable portion of time with the hand raised in an upper level of the workspace (within 40% of arm length of the shoulder elevation). With VERGE, for example, participants spent almost 20% of the session with their hand in this region of the workspace despite not having arm support.
Users, however, indicated differences in experience across the treatment modalities. While participants rated the modalities similarly in the weekly questionnaires, they expressed a preference for the HEP in certain areas in the comparative questionnaire, including as the most effective therapy and the treatment they would most likely continue in the home. Some of the appeal is undoubtedly attributable to the ease of use. Two-thirds of participants chose HEP as the easiest to use. This needs to be addressed in VERGE, as we describe in the following section.
Limitations and lessons learned
The study identified limitations with VERGE that need to be addressed for improved acceptance and utility. For example, the Trajectory Trace exercise placed a significant cognitive demand on users. They were required to actively cycle through a sequence of discrete states for each round (Draw, Claim, Trace, Reset) while coordinating with another player (i.e., one player would draw a trajectory while another would claim and trace it). It was sometimes difficult to determine the current state and to remember which came next. Thus, while almost equal numbers of subjects listed Ball Bump or Food Fight as their favorite VERGE exercise, none listed Trajectory Trace. Despite similar amounts of time spent on each VERGE exercise, hand displacement during Trajectory Trace was less than 50% of the amount seen during the Ball Bump exercise. Partially attributable to this, over 70% of participants chose HEP as the easiest to understand (only one subject picked VERGE). Clearly, reducing complexity of operation for the user of a therapy paradigm is of paramount importance. We have subsequently modified Trajectory Trace to include a visual display indicating the state flow and current state.
Due to the largely collaborative nature of the tasks in VERGE, they included limited quantitative performance measures for the users. Participants stressed the need for objective feedback. As noticeable functional changes may evolve slowly, quantitative assessment of game performance, which can show gains on a much shorter timescale, may provide the motivation needed in the short-term to enable reaching functional milestones. We have subsequently added scoring for each of the exercises. In some cases, both competitive and collaborative scoring is available.
There were limitations with the pilot study as well. While one of the potential benefits of the VERGE system is the inclusion of other players, we did not directly examine preferences for individual vs. partnered training, as previous studies have done [
26,
27,
29,
30]. In our study, the three training modalities were quite different from each other. The HEP and AWVR represented existing modalities with potential for use in the home. Factors such as ease of use, the engaging nature of the virtual task, or scoring undoubtedly influenced preferences for the chosen training modality. It should be noted that a large majority of participants expressed enthusiasm for playing with a partner when using VERGE and indicated that they felt that the presence of another user increased motivation. Enthusiasm for multiple players may have been even greater if a friend or relative had served as the playing partner, as was the case in a previous study [
26].
Our relatively small sample of participants displayed considerable motivation in repeatedly coming to the laboratory in the hospital for the study and maintaining study adherence. The enthusiasm for therapy may not be as great in the general population. Additionally, only three training sessions were performed with each modality. User responses may have been different after more sessions.
User response may also have been impacted by the fact that this pilot study was performed in the laboratory rather than the home. Coming to the hospital, interacting with people, and receiving compensation may have elevated interest, particularly for the HEP. Compliance rates for conventional home therapy exercise programs have been mixed [
53‐
56].
Conclusions
This represents one of the first tests of stroke survivors interacting with a remote user in a 3D virtual environment for therapy. The VERGE system can be directly utilized for home-based therapy with a family member or friend in their home or a therapist in the clinic. The low cost and minimal requirements make it practical for the clinic or home. Most participants expressed satisfaction with the system and enthusiasm for the virtual partner. However, they did stress the importance of ease of use and feedback of performance. Their responses highlighted the need for technology to be sufficiently flexible to accommodate the different goals and preferences of individual users.
Importantly, participants indicated a strong interest in home therapy. Over 66% responded that they would Definitely be willing to continue therapy in the home and 100% responded that they would perform the training at least 2–3 times per week. Two-thirds of participants indicated that they would be willing to perform home-based training 6–7 times per week. While limitations must be addressed, multi-user virtual reality environments hold promise for maintaining engagement in therapy and providing feedback of performance for home users. We are currently undertaking a home therapy study with the VERGE system.