Ethical consideration

The study was performed in accordance with the latest Declaration of Helsinki and received ethical approval from the Ethical Review Authority in Sweden (Dnr:2023-00353-01). All participants gave their written consent prior to participation.

Participants

Thirty subjects were recruited to participate in the study, whereof 28 were included in the final analysis (17 males, 11 females, mean age 25.3 years (standard deviation (SD) 4.6 years), mean height 1.80 m (SD 0.11 m), mean weight 75.8 kg (SD 13.5 kg), mean BMI 23.4 (SD 2.8). Two participants were excluded due to technical issues with loading stored data. Participants were included if they were young adults (range 18–45 years), considered themselves healthy, and had no prior history of diseases that could affect their balance, e.g., a known history of vertigo or neurological diseases. The participants were screened for health disorders by using a custom-made questionnaire, e.g., whether they had reduced balance, prior injury in the lower extremity the last year. The questionnaire has been used in previous studies (Almqvist Nae et al. 2024).

The recruitment of participants was conducted through contacts with students at Lund university, friends and family. Those who accepted to participate in the study were referred to the Movement and Reality Laboratory (MoRe-lab) at Lund University, Sweden, for the testing.

Electromyography and postural stability assessment equipment and EMG normalization data collection

The study hardware was designed to sample, synchronized in time, data at 1000 Hz from a posturography system and 2000 Hz from an electromyography (EMG) system. The posturography system was an AMTI force platform (AMTI HPS 464508, AMTI Europe GmbH, Helmstadt-Bargen, Germany). The EMG system was a wireless Noraxon Ultium system with 16 channels (USA. Inc, Scottdale, Arizona, USA), set to lowpass filter at 500 Hz sampled muscle activity before analysis. The skin was shaved and abraded using a medical abrasion gel (Nuprep, Weaver and Company, Aurora, Colorado, USA) before electrode placement. Dual surface electrodes with an inter-electrode distance of 20 mm (Noraxon, USA, Inc) were placed bilaterally on the postural muscles gluteus medius (GlutMed), medial gastrocnemius (MedGas), tibialis anterior (TibAnt), and peroneus longus (PerLong), according to the SENIAM (surface electromyography for non-invasive assessment of muscles) guidelines (Fig. 1 and Appendix 1) (Hermens et al. 2000).

Fig. 1

Electrode location for each muscle. A m tibialis anterior and m. peroneus longus, B m. gastrocnemius medial portion, C m. gluteus medius

Before performing study assessments, the maximum voluntary contraction (MVC) was assessed for each muscle, i.e., hip abduction for gluteus medius activation, plantar flexion for medial gastrocnemius activation, dorsi flexion for tibialis anterior activation, and pronation for peroneus longus activation. The procedure followed the SENIAM guidelines (Hermens et al. 2000), with a modification for gluteus medius MVC, which was performed in supine position instead of side-lying to prevent potential displacement or detachment of the bilaterally placed GlutMed electrode (Appendix 2). The MVC assessments were repeated three times for each muscle with 30 s pause between contractions. During each trial, subjects were instructed to gradually build up force and perform the contraction with maximal voluntary effort.

Test procedure

The preparations and performance of all tests were executed on the same day and required approximately 90 min per person. The participants performed seven postural stability tests. Two control tests (quiet stance with eyes open, quiet stance with eyes closed), were performed in a randomized order before five VR sessions. During the control tests, the participants were instructed to stand once with their eyes closed and once focusing for two minutes on a target positioned at eye level on the wall 5 m in front of them. They were instructed to stand on a force platform without shoes and with their feet positioned with a slight outward rotation (approximately 30 degrees) using guidelines on the force platform, and thus, to adhere to a standardized posturography test protocol (Fransson et al. 2019; Patel, Fransson, et al., 2009; Patel et al. 2020). The participants were instructed to stand in a relaxed position, with their arms folded over their chest, and to keep their face straight forward. The control tests were used for assessing the normal postural sway and to provide a relative reference for the stability recorded when the participants were exposed to distortive visual information from VR. The participants did not wear the VR headset when performing the control tests since the VR equipment precluded proper visual feedback.

Before performing the five VR sessions, participants were shown a winter landscape 360° video while sitting. This was done to calibrate the sharpness of the images displayed through the VR visor before performing the postural stability assessments. The participants were then instructed to stand in the same way as during the control tests, on a force platform. The Oculus Quest 2 VR headset (Facebook Technologies, LLC, Menlo Park, California, USA) used in the study tracks the head movements in 360-degree range, and thus, any change in head direction produced an identical direction change in the VR movie displayed to the subject. Performing substantial head movements in a VR environment may produce additional strain for the participants. Hence, to make the test conditions standardized and as similar as possible for all subjects, the participants were instructed to look straight ahead during the VR sessions. A Quest 2 VR headset weighs about 503 g, which corresponds to the weight of a standard construction helmet.

Five postural stability assessments were thereafter made while exposed to the VR stimuli, repeatedly watching a photographic 360° video of “a roller-coast ride”, where the ride included numerous quick turns to the right and left and ascending/descending elevations. The same video was shown for 120 s on each session, with a 10-minute rest period between sessions. The participants were allowed to sit down and rest between repeated VR sessions. Illustrations of the kind of postural control responses evoked by watching an immersive VR movie can be found in the paper Fransson et al. 2019 (Fransson et al. 2019).

Posturography analysis

The stability in anteroposterior and lateral directions during the control tests and repeated VR stimuli were determined by analyzing the variance of anteroposterior and lateral torque values from the force platform recordings. The torque variance values reflect the energy used towards the force platform to maintain stability (Johansson et al. 2009; Riccio and Stoffregen 1988). Moreover, biomechanical models of the human postural control system reveal that force platform recordings are directly influenced by two anthropometrical factors (i.e., the test subject’s weight and height) (Fransson 2009; Johansson et al. 2009). To appropriately address biases from these two anthropometrical effects, recorded torque values was regarded to reflect postural stability first after they had been normalized using the participants’ squared height and squared weight (Fransson et al. 2019; Johansson et al. 2009). The squared components in the variance formula made it necessary to normalize with squared parameters to achieve unit agreement.

In addition, spectral separations were performed to describe the smooth corrective changes of posture (i.e. low frequency energy < 0.1 Hz) and the fast corrective movements to maintain balance (i.e. high frequency energy > 0.1 Hz) (Kristinsdottir et al. 2001), using a fifth-order digital Finite duration Impulse Response (FIR) filter.

EMG analysis

EMG data for the test conditions and the MVC assessments of GlutMed, MedGas, TibAnt and PerLong muscles were first processed from artifacts (Karacan et al. 2023) by using a bandpass filter with cutoffs at 20 Hz and 500 Hz and then calculating the mean root square (RMS) envelope using a sliding window of 75 ms. The MVC values were thereafter obtained by filtering the EMG recordings, using a moving average filter with a sliding window of 1 s, and then selecting the highest value in the resulting data. The EMG activity during the different test conditions was attained by normalizing the EMG for each muscle during test conditions by dividing the filtered and RMS averaged data with the corresponding MVC values. The co-contraction index (CCI) was calculated on sample level for the three muscle pairs: (1) gluteus medius right and left side (GlutMed L vs. GlutMed R), (2) tibialis anterior and medial gastrocnemius (TibAnt vs. MedGas), and (3) tibialis anterior and peroneus longus (TibAnt vs. PerLong). For each muscle pair, the muscle with the highest normalized EMG activity was regarded as agonist. CCI was computed using Rudolph’s method (Rudolph et al. 2001) with the normalized EMG amplitude values:

$$\:CCI=\:\frac\left(antagonist+agonist\right)$$

In the statistical analysis, the mean and SD values for the CCI values were evaluated. These values were first calculated individually for each leg, Thereafter, the mean values using data for both legs were calculated and analyzed. The MATLAB 2023a software (The Mathworks Inc, Natick, Massachusetts, USA) was used for performing mathematical analyses.

Statistical analysis

A repeated measures General Linear Model (GLM) Analysis of Variance (ANOVA) analysis was used for evaluating the performance across repeated VR sessions. The independent main factor analyzed for VR sessions were: ‘Repetition’ (Session 1…5; degree of freedom (d.f.) 4)). In the analysis, the dependent variables were the log-transformed normalized total, low frequency, and high frequency torque variance values (posturography data) and the log-transformed mean and SD muscle CCIs (EMG data). The model parameter Repetition was a within-subjects factor.

The Wilcoxon matched-pairs signed-rank test (Exact sig. 2-tailed) was used for post hoc within-subjects analyses of the accumulated effects of repeatedly performing VR Sessions i.e., analyzing the adaptive changes between VR Session 1 and VR Session 5; the difference between VR Session 1 and the eyes open control test; the difference between VR Session 5 and the eyes open control test; and for determining the role of vision using data from the control tests (Patel et al. 2008).

A Spearman’s correlation was performed to establish whether associations existed between postural stability values and the CCIs. The correlations describing the characteristics during VR were based on pooling together the data collected from all 5 VR sessions. The correlations describing the characteristics during control conditions were based on pooling together the data collected from both control tests. In the correlation analyses p < 0.05 were considered significant and the R-values were evaluated according to Cohens thresholds: >0.1 = small, > 0.3 = moderate, > 0.5 = strong and > 0.7 = very strong correlation.

In the GLM ANOVA analysis, p < 0.05 was considered significant. In the post-hoc analyses, p < 0.025 was considered significant after Bonferroni correction of the within-subjects’ analyses of VR data and p < 0.05 was considered significant for the within-subjects’ analyses of control tests data. Non-parametric statistical methods were used in the post hoc tests as the Shapiro-Wilk test revealed that some datasets were not normally distributed, and that normal distribution could not be obtained by log-transformation (Altman 1991).

Sample size analyses of the posturography parameters revealed an effect size of 0.95 which shows that with the p-value set to 0.05 (2-tailed), our study requires 11 subjects to reach a power value of 0.8 for the parameter used. The statistical analyses were performed using SPSS (Version 28, IBM Corp, Armonk, NY, USA) and the power analysis was performed with GPower 3.1.9.7.