A total of 20 healthy participants (age: mean 45 years, SD 26 years) were asked to participate in this study after institutional review board approval was received. Physical health screening was performed on all participants to exclude anyone with balance issues, dizziness, or mobility deficits. After receiving informed consent from each participant, screening was conducted to exclude individuals with neurological and orthopedic disorders and any potential balance or dizziness issues.

The participants were asked to stand quietly, either with their eyes open or with their eyes closed, on a force plate (NeuroCom; Natus Medical Incorporated) while wearing the HTC Vive. Three 20-second trials were performed for the eyes open and eyes closed conditions. During the trials for the eyes open condition, a static virtual scene was shown to the participants to reduce visual feedback and to help them maintain their balance. Figure 1 shows the virtual scene (Unity version 2018.3.8; Unity Technologies) that was seen by the participants of this study.

The COP data were recorded by the force plate at a sampling frequency of 200 Hz. The VR position data were recorded at a sampling frequency of 10 Hz. The COP and VR position data were computed to find the normalized path length (NPL) [11], root mean square (RMS) [11], and peak-to-peak (P2P) value [11] for sway in the medial-lateral and anterior-posterior directions by using a customized MATLAB code (MathWorks). The NPL, RMS, and P2P value were calculated as follows:

In these equations, t is the time duration, N is the number of samples, p_avg is the mean of the data, and p_j is either COP data or VR data at time sample j.

The test-retest reliability of the NPLs, RMSs, and P2P values in the three trials for each test condition were examined by using intraclass correlation coefficients (ICCs) and 95% CIs. A 2-way mixed-effects analysis of variance was conducted with a model comprised of the subject and trial numbers. ICCs were classified as poor (<0.5), moderate (0.5-0.75), good (0.75-0.9), or excellent (>0.90) [12].

The strength of the association between the COP and VR outputs was computed via linear regression by using the NPL, RMS, and P2P value of the COP data as the dependent variables and the VR output as the univariate predictor for each measure separately. The data from the first trial and the 3-trial averages were analyzed. Coefficients of determination (R²) were reported as the amount of variance in COP data, which was estimated by using the VR outputs. The significance level was set to α<.05. IBM SPSS Statistics software (version 25.0.0.1; IBM Corporation) was used to conduct the statistical analysis.

The test-retest reliability of VR position outputs was similar to that of COP data for the RMS and P2P value. Our results suggest that the position outputs from the VR headset had a strong correlation with postural sway variables, such as the RMS and P2P value computed from the COP data. However, the NPL of VR outputs had a weak correlation with the NPL of COP outputs, and this might have been due to the characteristics of VR position data and COP measurements. Force plates take COP measurements from below an individual’s center of gravity, and VR headsets take position measurements from above an individual’s center of gravity. Therefore, postural sway in an individual would have had a greater impact on the headset’s measurements than on the force plate’s measurements, resulting in weaker correlations in NPLs.

A different brand of VR headset was validated for measuring balance, and the results showed good to excellent correlations between COP and VR headset outputs [9]. Our study demonstrated similar results for a different brand of VR headset. Moreover, our data further validated the test-retest reliability of the VR headset we used. The Wii Balance Board (Nintendo) was used to take COP measurements in a study by Marchetto and Wright [9]. The Wii Balance Board was designed as a video game controller with a low sampling frequency (40 Hz) [13]. In our study, a laboratory-grade force platform was used to collect the COP data. Although the Wii Balance Board was validated for taking COP measurements, a laboratory-grade force platform may provide more accurate data, especially in studies that validate other devices for balance assessments [13,14].

The average ICC was better in the trials for the eyes closed condition. This may have been due to the lack of variability in postural sway when standing with eyes open [15]. The ICC values may improve if variables with larger variations or more challenging balance variables, such as optic flow, are analyzed. Moreover, the sample size is another key factor that affects the ICC. Having a larger sample size may help to improve the test-retest reliability of VR position outputs.

This study presents promising results that indicate the usefulness of a VR headset as an alternative device for measuring balance control. Overall, the position data that were recorded by the VR headset correlated strongly with the COP data that were recorded by the force plate. The RMSs and P2P values of the data seem to indicate that there may be magnitude differences between the position data recorded by the headset and the COP data recorded by the force plate. Future work could be conducted to establish how much of a magnitude difference this is and if further data manipulation is necessary to obtain better correlations between the COP and position data.

Although the data show promise, there is still a need for improvement. In several trials, the correlation values were lower than 0.2. This indicated weak correlations among data sets. Since these correlation values tended to be outliers, it is likely that these lower values stemmed from calibration or procedure issues. The data analyzed in these trials also may have exhibited little variability, and this could have contributed to the low correlation values. Further software development for taking measurements by using the headset can be conducted, and improvements to the software used to analyze the data can be made. The further calibration of the headset and other data collection devices may also improve results. Adjustments can be made to the procedures used during data collection, such as syncing the headset and force plate to start data collection at the same time by inputting the same input command on a single PC. Future work can be conducted to investigate the correlations between position data recorded by VR headsets and position data recorded by motion capture systems. Comparisons between such data can be conducted to further support the use of a VR headset as a clinical and research tool.

Using VR position outputs could be an alternative way of measuring postural sway. However, a standard method will need to be established before such data can be used in this manner. Overall, VR headset position outputs appear to have good potential for being used in balance control studies. The lower cost of a VR headset system is an advantage and promotes the use of this device in clinic settings. However, further validation and software development may be needed.

An affordable home-based VR headset with a built-in gyroscope and triaxis accelerometer for tracking the position of a user was developed to promote video games. The goal of this study was to investigate the use of a VR headset as a cost-effective alternative to a force plate for measuring postural sway. The test-retest reliability of the VR headset and the postural control variables that were computed by the VR headset and a laboratory-grade force platform were compared. The lower cost of a VR headset system is an advantage and promotes the use of this device in measuring postural sway. However, further validation and software development may be needed.