A team of pain researchers and VR researchers created a simple open-source VR application for patients with CRPS using the game engine Unity 3D. This application is compatible with a range of consumer VR systems that use head-mounted displays and hand controllers or hand tracking to allow users to interact with virtual content by controlling avatars. Figure 1 shows an example of a user’s view of their avatar’s hands which are controlled by their movements.

As in conventional MVF, movement from the patient’s uninjured hand was transformed over the midline to symmetrically animate their avatar hand on the injured side. Thus, patients’ uninjured hands could control the movements of both avatar hands, just as in conventional mirror therapy.

In the summer of 2017, 110 adult patients in the database of a pain clinic at a large United States university with a diagnosis of unilateral CRPS in an upper extremity were contacted to ask if they would be interested in participating in a pilot study to test the feasibility of a novel VR intervention. Eligible patients had average pain level reported on a Visual Analog Scale as greater than or equal to 4 out of 10, were on a stable treatment plan for the month preceding the study, and could commit to multiple visits. A total of 9 patients (6 female and 3 male) were recruited within the 2-month time span set for the study. Among the 9 patients, 7 (78%) identified as Caucasian, 1 (11%) as Asian, and 1 as “other” (11%). Their ages ranged from 19 to 60 years, with the median being 42 and mean 44. Their education level ranged from a high-school diploma to graduate degrees, and their income ranged from under US $10,000 to US $80,000 or more. All 9 patients were right handed, with 3 having CRPS on the left upper extremity and 6 with right upper extremity CRPS. All patients signed informed consent in the clinic, received US $20 per hour of visit time, and the experiment was approved by the Stanford Institutional Review Board. The consent form, along with all other study materials, is shown in Multimedia Appendix 1.

Table 1 shows participants’ symptom onset and duration.

To make this environment as accessible to as many users as possible, it was built to accommodate both the hardware systems currently supported by the Steam VR plug-in in the Unity game engine. At the time of the experiment, these were the Oculus Rift/Touch system [14] and the HTC VIVE system [15], both of which cost less than US $400 and could be deployed using a “VR-ready” laptop as seen in Multimedia Appendix 2. The application can accommodate, but does not require, room-scale tracking. Since this initial study, the project has been revised to be used with standalone headsets that include hand and finger tracking. Original and revised projects and documentation can be found in [16].

All visits took place in July and August of 2017. After the screening interview, participants completed a baseline survey at the clinic through the institution’s REDCap account. This survey consisted of a set of demographic and baseline measures, including the PROMIS measures of pain and daily function [17]. This survey was repeated after the last session, and 1 month after the study. All survey instruments are provided in Multimedia Appendix 1.

Participants then returned to the clinic weekly for a minimum of 4 sessions of immersive VR, with an optional fifth visit based on participant interest and availability. After each session, participants completed a daily pain survey (modified from the Brief Pain Inventory [18]) and a presence questionnaire derived from Witmer and Singer [19].

Participants were seated in a swivel chair in front of the computer and a researcher (AB) helped them to don the headset and hold the hand controller in their unaffected hand. Avatars were scaled to the participants’ seated height and the color of the arms and hands was selected at the beginning of each session. For participants suffering from CRPS of the right upper limb, both avatar hands were controlled using the left controller. For participants suffering from CRPS of the left upper limb, this was reversed and both avatar hands were controlled using the right controller. It was too painful for some participants to hold the tracker with their injured hand, so all participants held only 1 tracker on their uninjured side.

The researcher (AB) then began the session by pressing the spacebar key. Each time the key was pressed, one of a series of 18 targets of 3 different sizes (6 large, 6 medium, and 6 small) appeared in random order in the midline of the participant’s field of view. Participants were asked to bring both the injured and uninjured hands together to contact the target. When an avatar hand contacted the target, a chime sounded, and the target object disappeared. Participants completed as many sets of target hitting as they wished (mean 7.5 [SD 2.25]). For the first 2 participants, the protocol differed slightly for their first 2 visits, in that they were only allowed to complete 2 sets.

The results of the baseline surveys can be found in Multimedia Appendix 3.

Several participants commented positively on the sessions, for example “VR helped hand pain!” and “I’m more likely to use my affected arm than I was before the study. Even my family noticed.” However, using a linear mixed model with Bonferroni-corrected α of .01 for multiple comparisons, there were no statistically significant differences over time on average or highest pain of the affected limb or body, or on physical activity, mood, or quality of sleep. Table 2 shows these results.

Most consumer VR systems contain trackers on the head-mounted display and 2 hand controllers, each of which captures both position (X,Y,Z) and orientation (pitch, yaw, roll) data at each frame. The default setting in our current module is to record every third frame, for a rate of approximately 30 frames per second. However, the frame rate can be modified to record at a higher or lower rate. We saved movement data for each tracker to a separate CSV file at the end of the experiment. We note that the data we recorded reflect the actual positions of the single tracker held by the participants, and the head-mounted display that they wore on their heads.

In this pilot study, participants’ head and uninjured hand movements were tracked. Participants were not asked to hold or attach a hand sensor to their injured hand, on the grounds that this might exacerbate their injury. Thus, the data collected from these sensors cannot inform us about how participants may or may not have moved their injured arm during the experiment. Data collected from the other 2 sensors can be used to examine changes in movement over the course of the following sessions, and can also be compared with other, self-reported measures. A representative example of movement data can be found in Multimedia Appendix 4.

The majority (7 participants) reported no cybersickness, with 2 participants rating their cybersickness as “slight.” No participants quit any trial due to its effects.

In this pilot study, we tested the feasibility, ease of use, and patient willingness to engage with an open-source MVF VR module for CRPS. All 9 participants were able to use the system for 3, 4, or 5 trials each, all participants stated that they were interested in an at-home trial if one were made available, and no participants stopped any trial due to cybersickness. We did not find a statistically significant effect of trial on participants’ self-reported average pain scores, but this is not surprising given that the study was not intended or powered to detect a clinical signal for efficacy.

Because of the small number of participants, we cannot draw conclusions about efficacy nor do we have enough information to discuss generalizability. In addition, participants were not blinded to treatment, and the researcher working directly with the patients (AB) was also not blind to the purpose of the study.

Because in this study, participants only held the trackers in their uninjured hand, we could not directly track improvement in movement on the injured side.

Although hand tracking was not available at the time this study was conducted, we have since updated the project to include markerless hand tracking, including finger tracking through the Oculus Quest. This can facilitate tracking of the injured hand, and may also improve the overall patient experience. Future work should examine whether markerless tracking and finger tracking improve patient outcomes.

Other potential easy modifications include changing the appearance of the avatar body, changing how the avatar limbs are controlled, and changing the appearance and placement of the targets. We have made this project open source with the goal of encouraging further exploration of deploying MVF in immersive VR using consumer systems.

This pilot study measured participant pain scores using a daily pain survey modified from the Brief Pain Inventory [18]. Future studies may use other standardized pain indicators, for example, Pain Self-Efficacy Questionnaire (PSEQ) [20], EuroQol-5D (EQ-5D) [21], or Pain Catastrophizing Scale (PCS) [22], as well as analyzing improved function as an indicator of positive outcome.

All 9 participants in our initial pilot study were interested in an at-home trial if one were made available. While at-home VR systems are still out of reach for most consumers, such systems can increase accessibility [23]. An at-home module would allow patients to use the module at will and could also be a useful channel of communication between the clinician and the patient. In such a system, safety and social interaction must be fully considered. Thus, any consent form for such an environment should reiterate potential risks and protection of patients’ data must also be assured [24].

We note that this software is not a commercial system or patient-ready application. Rather, it is designed to be used as a potential starting point for research for other clinicians interested in exploring potential therapeutic uses of immersive VR for visual feedback on movement. We look forward to the contributions of other researchers in this area.