Chronic pain (CP) is a common condition occurring in 1 in 5 Canadians [1], and it has limited effective treatment approaches. Mindfulness-based stress reduction (MBSR) has shown some evidence of efficacy in this area [2,3] and has also been used to treat other clinical conditions, such as migraine, depression, addiction, and substance misuse [2,4].

Mindfulness meditation encompasses a range of mental exercises that share a common focus on regulating attention and awareness to improve well-being [2-4]. It is described as the quality of being completely engaged in the present moment, free from distractions and judgments, and being aware of bodily sensations, thoughts, and feelings without getting caught up by them, and it is used as a therapeutic technique in MBSR [5,6]. These practices involve mental training that allows practitioners to develop their minds in specific ways to help them deal with stress and anxiety [7,8]. Although the clinical benefits remain somewhat controversial, it is generally viewed as a beneficial practice for mental well-being, stress reduction, and pain management [9-11].

A recent trend in MBSR practice has been the use of immersive virtual reality (VR) to help participants focus on meditative exercise [12-16]. However, to date, neurological studies have not been performed with VR-guided meditation practices. Therefore, this exploratory study sought to identify any identifiable neurological effects of VR-guided meditation practices using electroencephalographs (EEGs).

Subjective reports of the benefits of mindfulness meditation have prompted investigations into the potential corresponding neurophysiological states. Exploration of fluctuations in brain wave voltage amplitude (power) topography and coherence (associated areas of activity) using EEG variations in neural activity assessed with functional magnetic resonance imaging and cortical evoked responses to visual and auditory stimuli that reflect the impact of meditation on attention [4,17-26]. However, findings remain speculative. EEG studies have previously reported modulation in α, θ, and γ band frequencies, generally with increased power and coherence during mediation [4,7,17,27]. Some studies theoretically conjectured how meditative states relate to EEG bandwidths. For example, Travis and Shear [28] suggested that focused attention (sustained attention is focused on a given object) increases γ band power, open monitoring (nonreactive monitoring of an ongoing experience) increases θ, and an automatic-self-transcending stage (transcending the procedures of the meditation) increases α [28]. Nevertheless, a consensus on what we currently know about how EEG forms correlate with meditation or how this may map onto stages of mental development or specific cognitive skills is yet to be reached [7,8].

The α frequency band lies between 8 and 12 Hz and is predominantly located in the occipital cortex. α waves are present in deep relaxation and sleep, usually when the eyes are closed. θ waves are characterized by oscillations in the 4-8 Hz band found in both cortical and subcortical structures. Increases in θ have been described during a variety of learning and recognition tasks, light sleep (including the rapid-eye-movement dream state), and deep meditation. θ and α power changes have been reported to increase in a number of meditation studies [4,17,19,29-31]. Moreover, γ is a higher frequency range generally regarded as between 30 and 50 Hz, although the range reported has varied substantially between 20 and 200 Hz across different studies [30]. This initial research suggests that γ is associated with a number of sensory and cognitive high-level information processing functions, such as semantic insights, learning, and neural plasticity. Peak γ frequencies around 40 Hz in bilateral hemispheres have only been observed in highly practiced meditators [7,30]. In addition, a posterior increase in γ activity may be related to enhanced perceptual clarity reported in some open monitoring (focusing on awareness itself) meditative processes [7,17,30].

VR is rapidly developing as a new form of media and uses computer-generated audio-visual displays and hand controller user interfaces to produce a sense of immersion in a digital 3D environment. Instead of watching an image on a typical computer or video display, VR technologies provide an increased sense of presence by engaging the senses (sight, sound, and touch) in real-time stereoscopic audio-visual media where users can move around and explore the environment as if they were there.

One of the most common health care applications of VR is its use in pain management. Several studies have explored the value of MBSR in CP management, although the reported effect sizes for this technique have been typically mild to moderate, and regular adherence with meditative practice is problematic [3,32,33]. It has been suggested that combining mindfulness meditation within a VR intervention may help support acceptance and adherence to practice while having a synergistic effect on pain reduction through immersive VR distraction [34-36]. This remains an active area of research, as adjunctive VR strategies have been used successfully in the treatment of acute pain [37-41] and more recently have also been explored with CP [12,35,42,43]. The theoretical rationale behind why VR may enhance mindfulness skills is that VR provides the user with cognitive displacement by actively engaging in a coping activity that provides a profound sense of through presence in another world. Cognitive distraction is a common strategy in pain control and relies on creating competition for cognitive resources, that is, attention to a novel spatial orientation, and engaging within it reduces the perception of pain [32,33,44]. Therefore, immersive VR interventions using stereoscopic head-mounted displays (HMDs) have been proposed as powerful tools for providing visual, audio, cognitive, and emotional engagement [45-47]. In MBSR, VR experiences are typically accomplished using computer-simulated environments, stereoscopic headsets, and motion tracking to support a more immersive meditative experience. This was the approach taken in this study, and an EEG analysis was selected as a practical technique to assess neurophysiological activity while experiencing VR.

Overall, this inductive exploratory study sought to assess how EEG power of waveforms, their topographic mapping, and coherence measures altered in 3 main states during a VR-guided meditation experience in patients with cancer-related CP: at baseline (pre), during VR-guided meditation (med), and after VR-guided meditation (post). Moreover, we explored whether their pain level was associated with waveform power measures. We were particularly interested in the power, topography, and coherence of α, γ, and θ wave activity, and possible synchrony with VR activity, as other researchers have reported changes in these 3 waveforms with MBSR activities. Therefore, we were interested in determining whether prior findings were consistent with those observed during a VR experience. Specifically, the questions we sought to address in this study were as follows:

An exploratory, single-subject design study was undertaken to compare EEG activity and pain levels before, during, and after VR-based meditation practice. This study was reviewed and approved by the University of British Columbia Clinical Research Ethics Board.

A convenience sample of 10 participants was used and recruited from those in an existing randomized controlled trial (RCT; ClinicalTrials.gov: NCT 02995434) where patients with cancer were using VR as an adjunctive therapy to help manage their CP (Textbox 1). These participants were completing or had completed cancer treatment and experienced a range of cancer-related pain, including neuropathy, fibromyalgia, postsurgical pain, or an exacerbation of pre-existing pain.

Participants were purposively recruited, focusing on recruiting those from the RCT cohort who had previously responded well to a VR-based meditation experience, with a self-reported reduction of a Visual Analog Scale for Pain ≥1. They were invited to participate in a single 2-hour VR-guided meditation experience, with EEG recorded in their home or at the university, and were offered a Can $100 (US $83) honorarium and expenses for taking part. As an exploratory study designed to establish methods and feasibility using a limited convenience sample, with an unknown effect size, the power to inform the sample size was not calculated a priori. This is acceptable for this type of inductive study [48].

EEG signals were recorded during the session using a BioSemi ActiveTwo system (BioSemi) with 64 channels in a standard 10-20 configuration. This system uses a head cap system with pin active silver chloride electrodes. The EEG ground (labeled DRL [Driven Right Leg]) on the Biosemi system was placed between POz and PO4, and the EEG reference (labeled CMS [Command Mode Sense]) was placed between POz and PO3. The antialiasing filter was a fixed first-order analog filter with −3 dB at 3.6 kHz, and the low-pass filter was a fifth-order cascaded integrator-comb digital filter with −3 dB at one-fifth the sample rate. A powerline notch filter was not applied because the system used active electrodes, a battery power supply, and optic fiber, greatly reducing noise from the powerline. Each channel consisted of a 24-bit analog-to-digital converter. Recordings were made at 1024 Hz (although one was recorded at 2048 Hz) using ActiView software (Biosemi Instrumentation).

An HTC Vive VR system (HTC) with a Deluxe Audio Strap fitted over the top of the EEG cap was used. This system features a 2160×1200 resolution, 90 Hz refresh rate, and 110° field of view. Positional tracking from 2 infrared cameras enabled 5-degrees-of-freedom motion tracking of the headset and hand controllers. Integrated stereo headphones supported 3D audio immersions. During the initial pilot testing, this system was found to work effectively. Minor noise was evident in the EEG from the VR system on rare occasions, but we were able to remove most of this noise by careful repositioning of the equipment.

For meditation practice, a commercially available Guided Meditation VR application (Cubicle Ninjas) was used. As each meditation was only 10 minutes in this app, we selected a single sequence of 3 unique meditations, Zen 2-4, to form a 30-minute block of meditation. In the guided meditation experience, users were situated in the Lost Woods virtual environment with the Calm music selected. They were able to explore a calm, forest 3D environment with running water features with soft chirping bird and gentle wind sounds. Using a controller, participants could move to different positions in the forest to explore or find a particular viewpoint they liked and found most conducive to their meditation. A narrative provided audio guidance on the meditative practice. This environment was selected to maximize the similarity with the participants’ prior VR experiences in the RCT.

Both EEG recording and the VR system were run on a Dell G7 17 7790 gaming laptop (Intel Core i7-8750H, 16 GB DDR4, RTX 2060) placed in front of the participant. During recording, the laptop screen was arranged to display the ActiView software and a mirror window of the Guided Meditation VR application experience, showing the participant’s perspective in the VR HMD. In addition, a Windows 10 (Microsoft) camera app was used to record the video. This arrangement was recorded using Snagit (Techsmith) screen capture software to accommodate the time synchronization of EEG and recordings of participants’ physical movements (Figure 1).

Participants were seated in front of the laptop, and the laptop webcam was framed to capture the participants’ head, arms, and upper body. Before putting on the EEG head cap, participants were familiarized with the VR app, its controls, and how to navigate and select meditations. The Cz, inion, and left and right preauricular locations were marked using standard EEG landmarking methods. These locations were used to align the EEG head cap on the participant. The electrode paste was then applied, and the electrode contacts were adjusted until the electrode offsets were less than 50 µV. The VR HMD was then placed in position over the EEG head cap, and the experience commenced.

The overall structure of data collection used an 8-minute resting condition, followed by the 30-minute sequence of meditations and followed by another 8-minute rest condition (Figure 2). The 8-minute period was considered a balance between being shorter to reduce strain and being longer for more data; this period has been used in EEG studies in a variety of fields for baselines [49-51].

During the rest conditions, the participants were asked to rest quietly and observe a small white crosshair displayed on a black background in the VR HMD. Participants were instructed to keep their eyes open while blinking naturally, to keep their eyes on the crosshair, and to stay still while not thinking about anything. During the meditation practice, participants were instructed to engage with the guided meditation and look around or move around the virtual environment as they liked, to find an enjoyable perspective. Each guided meditation experience lasted 10 minutes, and participants were instructed to begin the next one immediately after the intervening rest condition. Pain was assessed before and after the first rest condition, after each 10-minute guided meditation, and after the second rest condition. Participants were asked to verbally rate their pain from 0 to 10 on the simple numerical rating scale (NRS) [52].

A total of 10 participants were recruited (Table 1). As we went through the data cleaning process, the data from 1 participant (S05) were found to be excessively noisy, and on video review, 2 other participants (S08 and S09) were found not to be following the procedures. Hence, their results were excluded, and a sample size of 7 was used in the final analysis.

Figure 4 illustrates the average of the power spectrum of the 3 conditions to provide a general overview of the power spectrum in the data, whereas Figure 5 shows the box and whisker plot of band power of the pre, med, and post conditions in the 5 traditional frequency ranges. From these figures, the differences in power in the frequency bands between the conditions were identified. The post condition power level had the lowest median and mean values in the δ range. The med condition changed from the lowest median and mean values in θ and α to the highest median and mean values in β and γ.

The topographic distribution of the PSD is shown in Figure 6. Differences in pre, med, and post conditions are shown spatially in all bands and in different brain regions. Changes in the power levels during meditation were observed in all frequency bands. In δ, an increase in power level in the central occipital region was observed, and a drop of power in the frontal-central region was observed in the post condition. In θ, there was a drop in the power level in the frontal cortex. In α, a decrease in power level in the central parietal region was noted. In the β band, an increase in power level was found in the bilateral central and prefrontal regions during meditation. In γ, an increase in power level was noted in the left frontal (LF) and right frontal (RF) regions, and a slight increase in the central parietal region was observed.

Figure 7 shows the topoplot of the two-tailed cluster-based permutation test results. The conditions were compared in the test, including two-tailed pre versus med, med versus post, and pre versus post and a one-tailed MANOVA for all 3 conditions together. The clusters of electrodes with significant differences found in PSD changes were marked with a circle marker at the electrodes. The color bars indicate the t values computed from the test. For the pre versus med comparison, a cluster with a significance of P=.02 was found in the range of 24.5-31 Hz (high β and low γ) in the frontal cortex. For the med versus post comparison, significance was found in the β and γ ranges. A cluster with a significance of P=.001 was found in the range of 37-50 Hz (γ), and the cluster covered most brain regions; a cluster with a P value of .008 was found in the range of 23-36 Hz (high β and low γ) in the frontal and central cortices. For the pre versus post condition comparison, a cluster with P=.02 was obtained in the frontal, central, and parietal regions in the range of 8-9.5 Hz (high θ and low α). The MANOVA cluster-based permutation test for all the pre, med, and post conditions returned 4 clusters with P≤.05, where the test was one-tailed. The first cluster found has P=.002 in the range of 37.5-50 Hz (γ) in the frontal, central, and parietal cortices. The second cluster with P=.03 was found in the range of 31-36 Hz (low γ) in the frontal and central cortices. The third cluster with P=.03 was found in the range of 2-5 Hz (δ and low θ) in the frontal, central, and parietal cortices. The last cluster with P=.04 was found in the range of 24-30 Hz (high β) in the frontal and central cortices. Table 2 shows the test results for all the clusters of significance. The very large effect size indicated that the permutation tests were effective in rejecting the null hypothesis.

By inspecting the topoplots (Figures 6 and 7), 5 peak channels with noticeable changes in the power levels between the conditions were selected for additional analysis and for better understanding the observed power level changes. The peak channels selected were AF7 and Fp2 in the prefrontal region, FC1 in the frontal region, CP5 in the left central (LC) region, and P5 in the left parietal (LP) region. Two-tailed, paired-sample t tests were performed to examine the overall PSD changes in different conditions. These tests were conducted on the pre and med, med and post, and pre and post conditions. The resulting P values were reported with an FDR adjusted to 10%. The t test results were plotted graphically for ease of interpretation. For each of the peak channels, a combined plot of the results of the 3 conditions was used.

Figure 8 shows the P values obtained in a graphical form for the comparison of the 3 conditions. The upper shaded areas represent P≤.025. The shaded areas above the dashed line represent the adjusted significance with P≤.0025.

A statistical power analysis was also performed to verify the probability of detecting a true effect in the t tests. Table 3 shows the maximum effect size value and the respective frequency point found for each channel and for each comparison. Table 4 shows the statistical power based on the effect sizes listed in Table 3. It is shown that the t test results have a high power ≥0.7302 even when using the adjusted α value of .005.

Figure 9 demonstrates the coherence detected in the α band during meditation.

The functional connectivity level is represented by a red to blue color scale. The red color with a value close to 1 indicates that a channel pair is highly synchronized in the signal transfer. The blue color with a value close to 0 indicates that the channel pair is working independently.

Figure 10 shows the significant difference in coherence for the comparison of the med and post conditions in the α band. Channel pairs with P≤.025 are highlighted in green. Those with P≤.0025 (FDR adjusted) are highlighted in yellow. For clarity, the EEG channels were divided into regions according to their respective locations in the brain, namely, LF, LC, LP, left occipital, central parietal and occipital, RF, right central, right parietal and right occipital. Table 5 summarizes the results of a series of plots, as shown in Figure 10, for all the bands and all 3 comparisons. The table shows the significant coherence changes with P≤.0025 between the brain regions, where 1 denotes the pre and med comparison, 2 denotes the med and post comparison, and 3 denotes the pre and post comparison. The frequency band of the region pair at which a significance was found is shown with the Greek letter of the band. The number inside parentheses indicates the number of significant channel pairs in the frequency band.

As shown in Table 5, significant changes were mostly found between the frontal and parietal regions, namely, RF-LP (11 channel pairs) and LF-LP (4 channel pairs). In addition, the most active regions were LP (related to 17 channel pairs), RF (15 channel pairs), LF (7 channel pairs), right parietal (7 channel pairs), and central parietal and occipital (7 channel pairs). For the 3 comparisons overall, significant changes were mostly found between the frontal and parietal and occipital regions, particularly in the θ, α, and β bands in the med-post comparison. For the pre-med comparison, the frontal-parietal region-pair coherence changes were observed in the γ band. For the pre-post comparison, the frontal-parietal and occipital region-pair coherence changes were observed in the δ, α, and β bands. Table 6 shows the effect sizes of the 2 regions of interest, namely, RF-LP and LF-LP. Some large to very large effect sizes (≥0.8) were found in the δ, θ, α, and γ bands.

Table 7 shows the NRS pain scores collected immediately after the pre, Med1, Med2, Med3, and post conditions. A repeated measures correlation analysis was performed on some selected peak channels and clusters. In the test, no significant relationship between the collected pain scores and PSD was found.

As an exploratory study, the sample numbers were small and not necessarily representative of the wider population of patients with CP, which limited the power to identify differences. Therefore, there is a need for neurophysiological studies with larger samples to validate these results and to better explore this phenomenon. In addition, this was a single cohort study with no comparison group, although resting states before and after meditation were used as a no-mindfulness within-subjects control. Finally, the study focused on short-term neurophysiological alterations in the electrophysiology of the brain, and the long-term effects of meditative VR environments are still unknown, which will require longitudinal studies.

These findings suggest that distinct altered neurophysiological brain signals are detectable during VR-guided meditation, predominantly in terms of an increase in the power of the β and γ bands. Changes in the α and θ bands were also identified, predominantly as a pattern in VR-guided meditation compared with the resting baseline, possibly reflecting the specific impact of visual activity during VR-guided meditation. Some changes in coherence were also observed between the frontal and parietal and occipital cortices during VR-guided meditation. No significant association between NRS pain scores and changes in EEG signals was observed. Although this is an exploratory study, the results of this work clearly demonstrate the feasibility of EEG recording and subsequent data processing and analysis during VR experiences in patients using modern VR HMDs. To our knowledge, this is the first exploration of EEG alterations in the brain’s electrophysiological signals associated with VR-guided meditation in patients with CP and should provide some valuable initial data to inform future work in this field.