Virtual Environments for Vestibular Rehabilitation

Vestibular Rehabilitation Utilizing Virtual Environments to Train Sensory Integration for Postural Control in a Functional Context

Icahn School of Medicine at Mount Sinai, Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD)

The specific aims of this pilot project are: Aim #1: Determine the extent to which sensory integration strategies differ between 28 individuals with unilateral vestibular hypofunction and 28 age-matched peers. Participants' postural sway will be recorded as they experience two levels of moving stars10 and white noise, while standing on the floor or a compliant surface. Our working hypothesis is that patients with vestibular hypofunction utilize substitution strategies such that they will demonstrate greater visual and auditory reliance compared with controls, particularly when somatosensory cues are reduced via the support surface. We will then explore whether these mechanism changes after training. Aim #2: Develop the protocol and establish the feasibility of a randomized controlled trial (RCT) comparing C.S.I. training to standard vestibular rehabilitation. Following the assessment, the 28 patients will be randomized into standard vestibular rehabilitation vs. C.S.I. training. This pilot study will enable us to test the feasibility of our recruitment, randomization procedures, establish attrition rate, and test the training protocol. Aim #3: Generate pilot data for sample size calculation for a properly powered RCT. The follow up RCT will test the effect of C.S.I. training on: Visual Vertigo Analog Scale (VVAS), Functional Gait Analysis (primary); balance confidence, overall disability (descriptive). In our preliminary study, 8 patients met the inclusion criteria for the current proposal. Following the C.S.I. training, they had a large effect size of 1.17 on the VVAS. The current study will allow us to identify the between-group effect size for the VVAS and for a functional gait outcome.

System: The testing platform runs at 120 frames per second with either HTC Vive or Oculus Rift, on a Windows 10 laptop with 8GB RAM, Intel i7-7820HK CPU, Nvidia GTX 1080 Max-Q GPU, and Bose SoundTrue around-ear headphones II. The software was developed in C# with Unity3D 2018.2.0f1(64-bit) (Unity Technologies, San Francisco, California). The system utilizes SteamVR. Oculus Rift and HTC Vive both operate at 90 Hz refresh rate, 110 degrees field of view, and a high-definition video of 1080x1200 resolution for each eye. The graphics of the subway station, airport terminal building, and subway trains are modeled in Maya and imported to Unity3D. The rest of the 3D objects are modeled in Unity3D. The subway station model and airport model replicate a real subway station in New York City and a real airport terminal in the US. The contents in the system are fully controllable by the user interface (see Figure 3). Three-dimensional sounds were implemented using Wwise middleware and Google Resonance audio plugin. Using the head-tracking data, audio is modulated according to the position of the listener's head. The technology allows for creation of a rich soundscape in all directions around the listener. Audio assets used in the system are divided into two main groups: sound objects and ambiances. Sound objects are attached to the visual objects in the scene and their position is changing accordingly. Sounds include: footsteps, trains, announcements, cars, balls, airplanes, etc. Ambiances are created from original recordings from different locations in New York. These include different background sounds, i.e. sounds of the crowd chatter, distant trains, wind, birds, traffic, and general room tone of each of the spaces. All of the sounds used in the system are assigned to three different intensity levels which relate to the increasing complexity of the soundscape. Data Collection: The first assessment session will include a well-established postural control assessment that the investigators had translated to a HMD paradigm and added an auditory layer. The protocol will include all possible combinations of the following: 2 levels of visual perturbation (static stars; stars moving in the AP direction, 0.2 Hz, 0.032m); 2 levels of auditory perturbation (quiet room; white noise that cycles from 0 to 1 dB at 0.3Hz), and 2 levels of support surface (floor; memory foam). Each scene will be 60 seconds long, there are 8 combinations, and each will be repeated 3 times for a total of 24 trials. Power spectral density (PSD) of sway in 3 frequency segments will be derived from a laboratory force plate and will serve to explore sensory integration mechanism. Participants will also complete the FGA (primary), VVAS (primary), ABC (descriptive), and DHI (descriptive). Participants will complete the baseline assessment twice, one week apart, to assure stability of the measures. Following the baseline assessments, participants will be randomized to a C.S.I. experimental group (EG, n = 14) or a standard rehabilitation control group (CG, n = 14). The participants will commence the intervention program within a week of the second assessment. The setting will be a vestibular rehabilitation clinic at the New York Eye and Ear Infirmary of Mount Sinai. Both groups: Eligible participants will be provided with patient education and a basic home exercise program (gait, balance, no exercises with eyes closed) while they are considering participation in the study. Program Dose: 8 weeks, 1 visit per week, 30 minutes long EG: Progressive immersive training with the C.S.I. app; Scenes: start from most salient to the patient, eventually do all Duration: start at 60 seconds, increase over time up to 3 minutes per scene Complexity: start minimal, gradually increase up to most complex Tasks: standing with diverse base of support (BOS), head turns (progress with speed, planes); stepping, turning CG: Progressive gait, gaze stability and balance exercises. Gait: walking with head turns, progress with range, speed and planes of head movement; change of walking BOS: wide, normal, tandem Gaze: focus on a target while moving head side to side / up down. Progress with speed, duration, busier background, standing to walking. Balance: standing balance tasks, progress with BOS (wide to narrow to tandem), support surface, eyes closed, duration, head turns. Both Groups Progression / Regression rule: The highest level of challenge that can be done for 60 seconds with no loss of balance (LOB); No more than moderate symptoms in clinic based on the Simulator Sickness Questionnaire; If symptoms persisted over 2 hours post-session, scale the intensity back the next time. If symptoms improved immediately, repeat the task with the same intensity and duration. Home program for both groups: 8 weeks, 6 times per week, twice per day, 10 minutes long, Highest level of challenge that is safe (no LOB, no increased symptoms) for 60 seconds per task Home EG: Gait and balance exercises, No exercises with eyes closed Home CG: Gait, gaze stability and balance exercises, including exercises with eyes closed A post-assessment, identical to the baseline assessment, will be conducted within one week from the completion of the 8th intervention session. Sample Size: For Aim 1, the investigators will recruit 28 patients and 28 age and sex-matched controls. For Aims 2 & 3, the 28 patients will be randomized into an experimental (N=14) and control (N=14) groups. Assuming ~20% attrition during the course of the intervention, it is expected to have 11-12 participants per intervention group. Given our history of recruitment, this number is feasible for the duration and resources of the 12-month pilot grant. With that the investigators will determine the sample size needed for a future, adequately powered, study, as described below. Internal Validity: To avoid selection bias and to balance potential confounders, the investigators will use block-randomization strategy for participants using the blockrand package in R.21 This procedure uses a two-stage process: first, the size of the block is selected to 2, 4, 6, or 8; next, a block of that size is generated. Therefore, no researcher will be able to predict which group a participant will be assigned to or change that assignment (i.e. allocation concealment). Participants will only be randomized into groups after the baseline assessment and consent to participate in the study have been completed. Although, due to the nature of the intervention, participants and treating clinicians cannot be blinded to group assignment, bias will be minimized by blinding the post-treatment assessors to treatment status. To control for attrition bias, in addition to an intent-to-treat approach, the investigators will ascertain whether any predictors of missingness exist. If so, the investigators will use a pattern mixture model to multiply impute (5 replications) any missing responses. Consideration of Relevant Biological Variables: The eligibility criteria specified above represent adult individuals with unilateral vestibular hypofunction of both sexes and of diverse age range. The control group will be matched for age and sex with the vestibular group. A wide age range (18 and up) will be recruited. Obesity is known to affect posture and could alter postural control strategies. Should the sample include overweight individuals, the model will be tested while adjusting for weight. Because the sample will include aging individuals, the investigators anticipate additional health conditions that will be similar between groups, e.g., age-related hearing loss. The exclusion criteria will eliminate health conditions critical to postural control, such as: visual impairment, peripheral neuropathy, and other neurological conditions. Data Analysis: Aim 1: For each of the four measures of interest (PSD 1-3 and overall PSD), the investigators will fit a linear mixed effects model to compare the age-matched controls to patients with vestibular hypofunction while accounting for the inherent multi-level study design (person, conditions, repetitions). The models will include main effects of group, visual condition, auditory condition, surface condition, as well as their interactions, while adjusting for age. P-values for the fixed effects will be calculated through the Satterthwaite approximation for the degrees of freedom for the T-distribution. The analysis will be repeated following the intervention, adjusting for group assignment and pre-intervention values. ABC and DHI will be used to describe the sample. Aim 2: No statistical analysis needed. Aim 3: First, following the intent-to-treat principle, a linear regression model will be fit with the VVAS and FGA as the primary dependent variable, on treatment group, controlling for baseline covariates (age and other self-reported questionnaire measures) to improve the precision of the treatment effect estimate, as well as pre-test scores. The potential efficacy of the intervention will be assessed based on a significant coefficient for treatment status. Second, as described above, the investigators will fit a pattern mixture model to multiply impute any missing values, and use Rubin's rules to pool coefficients and their standard errors from the regression models fit to each imputed dataset. Data generated from this pilot randomized trial will be used to calculate the sample size needed for future, adequately powered, randomized controlled trials by estimating the difference in means for the treatment groups (i.e. a future expected effect size) as well as the variance of the primary outcomes and the values of other parameters necessary to compute the power function. Safety and Reporting: Preliminary work showed minimal to no cybersickness and no falls or adverse reactions to the virtual reality training. Only 1 patient dropped out due to concern about symptoms. Others dropped out due to anxiety, physician's request, cessation of therapy, or other orthopaedic injuries that occurred after enrollment. In the current study, Dr. Maura Cosetti, MD, will provide oversight and monitoring of the conduct of the trial as a preliminary DSMB. This will ensure the safety of participants and the validity and integrity of the data. The PI and Dr. Kelly will perform continuous monitoring of participant safety (falls, symptoms on every session) with frequent reporting to Dr. Cosetti. The investigators will follow the Consolidated Standards of Reporting Trials (CONSORT) 2010 statement checklist and diagram in reporting the results of the trial.

Progressive immersive training with the virtual reality app Scenes: start from most salient to the patient, eventually do all Duration: start at 60 seconds, increase over time up to 3 minutes per scene Complexity: start minimal, gradually increase up to most complex Tasks: standing with diverse base of support (BOS), head turns (progress with speed, planes); stepping, turning 8 weeks, 1 visit per week, 30 minutes long In home: Gait and balance exercises, No exercises with eyes closed, 8 weeks, 6 times per week, twice per day, 10 minutes long

Progressive gait, gaze stability and balance exercises Gait: walking with head turns, progress with range, speed and planes of head movement; change of walking BOS: wide, normal, tandem Gaze: focus on a target while moving head side to side / up down. Progress with speed, duration, busier background, standing to walking. Balance: standing balance tasks, progress with BOS (wide to narrow to tandem), support surface, eyes closed, duration, head turns. 8 weeks, 1 visit per week, 30 minutes long In home: Gait, gaze stability and balance exercises, including exercises with eyes closed, 8 weeks, 6 times per week, twice per day, 10 minutes long

The Visual Vertigo Analogue Scale (VVAS) is a self-reported questionnaire where a participant rates their visual vertigo on a 10 cm line in 9 different visually challenging environments. A score of 0 indicates no dizziness. Maximal score is 100 (calculated as the measurement on each item X 9 divided by 10).

A functional test designed to assess individual's ability to perform various motor tasks, such as: walking with eyes closed, walking backwards, climbing stairs. There are 10 items, each is scored by a therapist on a scale of 0 (severe impairment) to 3 (normal). Maximal score is 30. Higher is better.

The postural sway signal (derived from a force platform) is decomposed into its frequency components. This technique allows for quantifying overall variability of COP sway (COP variance = sum of power at all frequency components) as well as specific components associated with pathological and/or physiological balance control. In our paradigm, PSD 1 (0.05 to 0.25 Hz) includes the visual frequency and will indicate visual weighting. PSD 2 (0.25 to 0.5 Hz) includes the auditory frequency and will indicate auditory weighting. PSD 3 (0.5 to 1 Hz) will indicate small corrective adjustments, and overall PSD will indicate destabilization induced by the stimulus.

Inclusion Criteria: Adult patients (18 or older) Clinical diagnosis of chronic (3 months and longer) unilateral peripheral vestibular hypofunction Patients will be included based on the presence of a positive head thrust test, head shaking nystagmus, spontaneous nystagmus, and/or canal paresis > 25% if a caloric test is available. Included patients must present with at least two positive items on the VVAS Exclusion Criteria: Patients will be excluded for prior vestibular rehab, bilateral or unstable vestibular loss or another neurological condition, active benign paroxysmal positional vertigo, acute orthopaedic injuries, peripheral neuropathy, hearing impairment, or visual impairment not corrected with glasses.