A virtual reality head stability test for patients with vestibular dysfunction

1. Introduction

Visual dependence in patients with vestibular dysfunction has typically been defined by comparing patients’ postural sway between ‘eyes closed’ and ‘eyes open’ standing tasks [16, 18]. The sensory organization test (SOT) utilizes computerized dynamic posturography where the computer assigns a visual score to this difference in postural sway [17]. In a sample of 12 patients, and 50-second trials, Nashner found that patients with mild vestibular dysfunction were able to cope with the absence of sensory input, but not with conflicting sensory information [16]. Yeah et al. found that patients with vestibular dysfunction performed worse on the SOT during eyes closed tasks as compared with young adults, but not with older adults [20]. They argued that analysis of the frequency distribution made the SOT more sensitive, i.e., patients with dizziness showed significantly more high frequencies on all conditions with a sway-referenced platform as compared with healthy older adults. The applicability of the SOT to the clinical setting has been questioned, nor is it considered a diagnostic vestibular test [7].

The Clinical Test of Sensory Interaction for Balance (CTSIB) and the modified CTSIB (mCTSIB) [9] were developed as a simple clinical test based on a similar paradigm as the SOT where the outcome measure tracked is the time a patient is able to maintain a position up to 30 seconds [25]. CTSIB scores are intended to reflect the contribution of each sensory system to standing postural control [9]. Stance time on the CTSIB in patients with vestibular disorders was significantly reduced when compared with age-matched controls [4]. However, older adults and adults with vestibular disorders also demonstrated greater within-group variability as compared with young adults [4]. Mulavara et al. supported the usability of the CTSIB for screening of patients with vestibular dysfunction although their clinical group was older than the control group. They recommended adding simple kinematic parameters to the outcome analyzed, if possible [15].

The Balance Error Scoring System (BESS) was developed as a qualitative, sideline, sports rehabilitation alternative to the SOT [3]. It has been primarily used as a post-concussion assessment tool [5], however, norms have been developed for older adults [10] and adolescents [1]. To the best of our knowledge, the BESS scores of patients with peripheral vestibular dysfunction have not been reported in the literature [24]. Utilizing the more challenging BESS conditions, typically recommended for patients with Traumatic Brain Injury, could potentially be relevant and informative for patients with vestibular dysfunction who present a ceiling effect on the CTSIB or mCTSIB.

New technology may allow us to take a comprehensive and quantitative look at sensory integration for balance, while remaining portable and affordable and thus clinically accessible. Using the Oculus Rift headset, we implemented a novel test of sensory integration for balance via a short ‘moving room’ scene of stars [21]. Custom code, developed in our lab for the Oculus Rift sensor and accelerometers, was used to track participants’ head position and displacement in six planes of movement. The purpose of this pilot study was to compare balance performance of patients with vestibular dysfunction and controls on a subset of the BESS test and our new virtual reality (VR) scene. We hypothesized that a scene with mild visual disturbance (i.e., a moving room) will be more sensitive to between-group differences than traditional ‘eyes closed’ tasks. Our second purpose was to test the sensitivity of our new VR head tracking instrument as compared with a traditional laboratory forceplate.

2. Methods

2.3. Data processing

A Kistler 5233A forceplate recorded center of pressure (COP) oscillations at 100 Hz. We then applied a traditional low pass, 4th order, zero-lag, Butterworth digital filter using a forward & reverse pass with a cutoff frequency of 10 Hz. Because it has been shown that COP oscillations during static stance lie well below 3 Hz [2, 26], and our spectral analysis showed that our data met this prediction, our filter can be considered conservative. Head oscillations were recorded at 60 Hz using custom software to obtain the head position in 6 degrees of freedom (see Fig. 1), i.e., translation and rotation in Y (up and down translation and Yaw rotation), X (side to side translation and Pitch rotation) and Z (front to back translation and Roll rotation). For consistency, head oscillations were filtered in the same manner as the COP data. Using custom-made MATLAB (Matlab_R2017 for Macintosh) software, we calculated the following traditional measures of postural sway in the anterior-posterior (AP) and medio-lateral (ML) directions for COP data and in all 6 directions for head data. Directional path length (DP, mm for forceplate, cm for head translation and 0.01rad for head rotation) was calculated as the sum of the absolute value of position change at any time stamp [23]. Root mean square of velocity (VRMS, mm/s for forceplate, cm/s for head translation and 0.01rad/s for head rotation) was calculated as follows: [19] First, the difference in position between 2 consecutive data points is divided by the time interval (1/100 Hz), then the velocity at each point is squared and then summed. Finally, the square root of the sum is divided by the number of data points, according to the length of the time series (2000 in our case). Head data were only available for the virtual reality head stability test (VR_HST).

OPEN IN VIEWER

Fig.1

Degrees of freedom measured via the Virtual Reality Head Stability Test (VR_HST).

Each task was also scored clinically according to the BESS by a licensed physical therapist and a physical therapy student. A point is given for any error performed by a participant (0 indicates a steady, error-free trial). Possible errors include: opening the eyes, removing hands off hips, step, stumble or fall, lifting forefoot or heel off the ground, remaining out of testing position for >5 seconds. A trial where a participant could not hold the position statically for at least 5 seconds was scored as ‘fail’.

3. Results

Clinical BESS scores are displayed in Fig. 2 with each column representing the number of people per a given score across the 4 tasks in each group. All but one patient (who made only one error) scored a perfect 0 in the FL_FT_CL task. Controls had a perfect 0 score on the FL_TAN_OC task, whereas patients ranged from 0 to ‘fail’.

OPEN IN VIEWER

Fig.2

Distribution of the BESS clinical scores per condition for controls (top) and patients (bottom). Legend: FL FT = floor, feet together, eyes closed; FO FT = foam, feet together, eyes closed; TAN CL = floor, tandem, eyes closed; TAN OC = floor tandem with visual oculus display). Y axis: number of people that received a certain score.

Boxplots showing both groups’ COP DP per task can be seen in Fig. 3 (AP) and Fig. 4 (ML). While patients had longer DP on all tasks this difference was only significant for FO_FT_CL (p = 0.016 [AP] and 0.002 [ML]) and for FL_TAN_OC [AP] (p = 0.04) but not [ML] (p = 0.05). VRMS was also significantly higher for patients on these 2 tasks only with p = 0.019 [AP] and 0.001 [ML] for FO_FT_CL and p = 0.02 [AP] and 0.04 [ML] for FL_TAN_OC.

OPEN IN VIEWER

Fig.3

Boxplots comparing Directional Path (DP) in the anterior-posterior direction between controls and patients for the 4 tasks. Y axis is DP in mm. * indicates a statistically significant difference.

OPEN IN VIEWER

Fig.4

Boxplots comparing Directional Path (DP) in the medio-lateral direction between controls and patients for the 4 tasks. Y axis is DP in mm. * indicates a statistically significant difference.

Directional Path results from the VR_HST can be seen in Fig. 5. Overall, patients displayed longer head DP and more variability than controls (boxplots range is consistently larger for the patients). The directions that correspond to the forceplate, i.e., Tilt translation (AP) and Pitch translation (ML) were each significant at p = 0.019 for DP and p = 0.007 for VRMS and p = 0.01 for DP and p = 0.007 for VRMS respectively. Additionally, patients differed significantly from controls on all other head movement directions except for Yaw translation (moving up and down) where the P value was 0.06. As can be seen in Fig. 5, the differences were largest on all rotational movements, primarily Yaw rotation (p = 0.004 for DP and 0.003 for VRMS) and Pitch rotation (p = 0.007 for DP and 0.006 for VRMS).

OPEN IN VIEWER

Fig.5

Boxplots comparing head Directional Path (DP) between controls and patients derived from the Virtual Reality Head Stability Test (VR HST). DP (Y axis) is presented for Pitch (X) ,Yaw (Y) and Tilt (Z) translation and rotation. *Indicates a statistically significant difference. The units are meters and radians respectively, each rescaled by 100 (i.e., cm and 0.01rad).

In this study, we tested a novel assessment of visual integration for postural control via a short virtual reality head stability test. We compared the sensitivity of the VR_HST to that of a traditional forceplate and standardized static balance tasks, namely: standing with feet together or tandem on floor or foam with eyes closed. Our clinical sample included patients with peripheral vestibular dysfunction with large variability in severity of the condition. Both our clinical and control groups had a diverse age range. Accordingly, we observed great variability in performance, such that some patients swayed as little as the steadiest controls. Likewise, the control group had ‘outliers’ in performance that were closer to the median patients’ sway. This large variability in static balance and response to elimination of visual input supports the need to develop tools that could meet each patient at their level and direct treatment accordingly.

Clinically, we observed that simply asking patients to close their eyes while standing on the floor with feet together was easy for all patients (demonstrated by a perfect ‘0’ score on the BESS scale for all). At the same time, none of the patients could complete a 20-second-eyes-closed task in a single leg position (hence we could not record COP data for these tasks and eliminated them from analysis). However, tasks that demonstrated more diversity in performance, such as standing in tandem position with eyes closed, or standing on the foam with feet together and eyes closed were challenging for some controls as well. Interestingly, all controls scored a perfect ‘0’ on the VR tandem test and all patients improved their score compared with the tandem eyes closed performance. Even though the stars were moving, and all patients and controls perceived the movement, no patient felt sick or dizzy during or following the test. This suggests that VR test was challenging enough for these patients while not provoking symptoms. From a functional perspective, patients with vestibular dysfunction typically complain about dizziness or loss of balance in the dark and in open spaces. Therefore, a scene that mimics a dark, open space environment could be more salient to them than a complete elimination of vision by closing the eyes.

A portable instrument to quantify head oscillations and stability could be of great importance to patients with vestibular dysfunction. As can be seen in our sample, most controls had very little head movement during this static task. This was expected as a normal static posture response would be to dampen the head movement with low-frequency perturbation and utilize ankle and/or hip strategies [8]. This is not the observed pattern with our participants with vestibular dysfunction. It has been shown that individuals with vestibular dysfunction demonstrate poor coordination of the head during locomotion [22]. Lang et al. found that patients with vestibular disorders had a larger yaw rotation range when walking with eyes closed compared with healthy controls [11]. Indeed, during the VR HST, patients demonstrated greater head oscillations than controls in all planes of movement, a finding we could not observe clinically with a naked eye. This new VR paradigm, does not require 3-D video analysis or a marker-based system to obtain head position data. In addition, the portable head stability test could become an important tool for the assessment of patients with vestibular dysfunction during other tasks as well. Future studies could also quantify patients’ head stability using the head position sensor of the Oculus Rift during tasks with the vision obscured, by either trusting patients to close their eyes, or creating a completely dark scene.

The novelty of the tools and results demonstrated in this paper is limited by our small sample size, especially given the large variability of the clinical presentation of patients with peripheral vestibular dysfunction. There were more women in the patients group and lower body mass index (BMI) in the control group (potentially due to the gender distribution). In addition, while age was comparable between groups, the patients were slightly older. With larger samples, we will be able to start developing norms and classify patients with different levels of severities. Future studies should also explore the interaction between visual dependence and age, gender and BMI.

In conclusion, new virtual reality technology carries the promise of providing a tool with high levels of accuracy and precision. Utilizing flexibility in the code can accommodate clinical needs and provides a platform that is affordable and portable. Within this pilot study we detected significant differences in head control between a group of patients with vestibular dysfunction and a group of healthy controls despite large variability within each group and a relatively small sample size. We therefore hypothesize that large studies will demonstrate that our novel assessment will become an important component of a clinician assessment of the patient with vestibular dysfunction and potentially other patients with sensory loss leading to postural control deficits.