Healthy university-level participants were recruited from the community via email and word of mouth between October 2020 and January 2021. Exclusion criteria were recent neck pain, headache, concussion, cardiovascular instability (e.g., heart disease, high or low blood pressure), were actively undergoing cancer treatment, neurological or systemic conditions that affect balance or postural control (e.g., Benign paroxysmal positional vertigo), migraines, visual pathology (e.g., saccades), technophobia, and claustrophobia. This study was approved by a Health Sciences Research Ethics Review Board prior to participation in the study and all participants provided written informed consent prior to participation.

The VISION (Virtual Interface for Stress-Trauma Interactions through Open World Navigation) system is comprised of a HTC Vive (HTC Corporation, Xindian City, Taipei) head mounted VR display (HMD), noise cancelling headphones, and a Mikrolar R3000 (Mikrolar Inc., Hampton, NH, United States) robotic platform (Figure 1). The virtual environment these hardware peripherals interface with depicts a simulated car ride through a city, and the virtual reality environment was developed in the Unity game engine (Unity Technologies, San Francisco, CA, United States) (Figure 1). The visuals of this simulation (targeted at 30–40 frames per second) are delivered by the onboard screens of the HMD which also tracks the participant’s head position via two fixed infra-red base stations, allowing the participant to examine 360 degrees of their surroundings from the passenger seat of the moving vehicle. To maintain a heightened sense of immersion, the simulation’s audio is delivered by a set of noise-cancelling headphones such the volume of all external audio sources is reduced, further emphasizing the sounds of the virtual environment. The simulation is designed such that the movements of the participant’s virtual car are synchronized with a robotic platform which can be controlled along six programmable degrees of freedom: the x-,y-, and z-planes as well as yaw, pitch and roll. Prior to the collision, the robotic platform will simulate the feeling of acceleration and deceleration by modulating its yaw, pitch, and roll to match that of the virtual car. Instructions provided to the robot were designed to not accelerate beyond 1 g for any of the crash scenarios. On pilot testing we determined that for the rear crash scenario the peak accelerations at the head were no greater than 0.2 g. It should also be noted that the cervical spine regularly experiences such accelerations (1 g) in daily life without provoking symptoms (Allen et al., 1994; Castro et al., 2001).

The VISION system is capable of multiple different crash scenarios, through manipulation of the following crash parameters: the principal direction of impact (rear-end, side-impact, or front-end), time of day (night or dusk), weather (rain or clear skies), seat position (front passenger, or right/left rear passenger), audio selection (no music or various musical selections), whether or not the glass of the car shatters upon impact, and whether or not the participant receives audio indication that a collision is to take place (e.g., lights and horn of colliding vehicle). For this initial exploration of tolerability to the system, participants were exposed to a low amplitude rear-end collision (0.2 g) at night under clear weather while they sat in the front right passenger seat with no music playing. A night setting was selected for the time of day to help hide any lower fidelity graphical assets. The windshield glass of the virtual car was set to shatter and the participant received an audio indication that a crash was imminent via a simulated car horn played through the earphones. These parameters were selected as they were thought to be the least provocative in nature yet mirror those in rear-end MVCs, for which development of whiplash injuries has been reported to be the most common (Berglund et al., 2003; Dabbour et al., 2020).

Prior to their visit with the research team, each participant completed a study-specific demographic questionnaire indicating their sex and age. For the hour before the study, each participant was advised to refrain from any eating, drinking, or physical activity. After having their cervical spines screened, and 5 minutes of quiet sitting, each subject sat in the VISION simulator (Figure 1) during a non-crash route for 5 minutes to become acclimatized to being in a virtual reality simulation and to judge the immersion of the simulator. This non-crash route involved sitting in the front right passenger of a virtual car with it making four turns around a city block with no crash. Seat height of each participant was optimized to their respective height based on personal preference, the time of day was set to night, and no in-simulator music was playing. Each subject was allowed to survey their virtual environment as much as they wished during the virtual drive. Following the 5-min acclimatization period during the non-crash route, subjects were given a minute of rest prior to the next simulation. This rest period consisted of removal of the VR-headset, but participants remained seated in the robotic platform. Immediately following the non-crash simulation, each participant was informed that the subsequent simulation would include a virtual car crash (Figure 1) but were not told when or where (i.e., rear-end collision) it would transpire. Each participant was then exposed to the low-amplitude rear end collision through the VISION platform, that occurred after approximately 1 min of simulated riding in the virtual car. During the crash the robot delivered a small anterior perturbation (0.2 g) to the participant through the car seat to further the immersion of the simulation. Afterwards, the participants dismounted from the simulator, and were observed for 5 minutes to monitor for the presence of any immediate adverse reactions.

Descriptive statistics, including mean, standard deviation and range, were calculated for participant demographics. Data were visualized via box-and-whisker plots and outliers were identified through visual (in reference to interquartile range) and statistical tests of normality. The majority of the measures were non-parametric in nature (Kolmogorov-Smirnov tests). Individual items from the PQ and SSQ were critically examined independently to determine the usability of the VISION platform and to ascertain areas for improvement. SSQ items were recorded as a percentage of the total sample to understand incidence for each of the scores (i.e., none, slight, moderate and severe). As the PQ items do not have a descriptor for each part of the scale, individual items were examined descriptively (mean ± standard deviation). Adverse reactions were recorded throughout the protocol as incident counts of reported neck pain longer than 48 h, immediate vomiting, inability to tolerate the simulator, and self-reported disorientation post-protocol.

As this research project was exploratory in nature, no formal sample size calculation was performed, though a small sample of as little as 20 participants has been suggested as appropriate for exploratory studies (Daniel, 2012). A sample of 25 participants was deemed adequate to estimate tolerance to the VISION simulator in otherwise healthy young adults and to identify potential areas for improvement while minimizing risk of exposure to this potentially provocative scenario.

25 healthy participants (16 male, 9 female) with no recent (3 months) significant trauma or injury that required medical care participated in the study. The mean age of the group was 27.3 ± 4.1 years. All subjects successfully completed all aspects of the protocol and there were no verbal reports of nausea, inability to tolerate VR, and complete disorientation in response to the virtual simulation. There were no reports of neck pain immediately after the simulation. One subject did report neck pain on the day immediately after the virtual simulation, but their symptoms only lasted 1 day. No medical follow-up was required.

Results for the Simulator Sickness Questionnaire and Presence Questionnaire and their subscales (mean ± standard deviation, range) are presented in Table 1. In terms of Simulator Sickness, the mean total score was 23.49 ± 21.98 (0.00–89.76) out of a possible 235.62 indicating sufficient discomfort. The majority of participants indicated none to slight for each item of the SSQ items with twelve participants reporting moderate severity of a number of symptoms including nausea and fatigue, and only one participant indicating severe symptoms (eye strain) (Table 2). Mean PQ score was 91.04 ± 14.08 (54.00–112.00) out of a possible 133. Participants rated the ability to control events (mean 2.04 out of 7.00), environmental responsiveness (mean 3.40 out of 7.00) least favorably compared to the remainder of the PQ items (mean 5.00 out of 7.00; Table 3).

The sample size of this exploratory study was small and as such its conclusions must be interpreted with caution. However, this study sought to seek initial responses to a novel virtual car collision simulator and sought to explore initial reactions and the presence of adverse reactions. As such, we believe we were successful in this endeavor and that the results of the study will help to inform larger study designs to answer additional questions. Another limitation with this work is that we did not control for multiple comparisons as the work was preliminary in nature. However, we feel that due to the exploratory nature of the research the results still hold merit and invite future investigation. As mentioned earlier, we also did not control for subjects with previous virtual reality experience, nor did we control for participants who had been involved in previous car crashes. Both variables could influence the sense of presence and/or simulator sickness in response to the VISION platform. While this study explored the initial usability and tolerability of the VISION platform, formal interviews with future participants may also help to provide further feedback for optimization.

Additionally, there were rare occurrences of the base stations losing track of the participant’s head in 3-D space. As a result, the system had to be re-set to re-track the participant’s location. These tracking errors may have led to lower feeling of presence during the simulation. Again, with further refinement, it is expected that the incidence of these episodes should be greatly reduced. The framerate of the VISION system also requires optimization, as framerates were found to dip below 30 frames per second at times. Lower frame rates could partially explain feelings of simulator sickness. Participants were also made aware that a collision would occur which may have influenced their reaction to the simulator. The VISION platform also only allows for participants to be passengers as opposed to drivers of the virtual vehicle. Examination of the experience from the driver’s perspective is an area for future development, and the same hardware has been used for simulated forklift driving (Dickey et al., 2013). Lastly, we only used a small sub-set of the VISION platform’s settings. As such, it is unclear how the other settings (e.g., direction of impact, in-simulation music, awareness of impending collision) may impact stress, the sense of presence, and simulator sickness.