Chronic pain is a major global health care problem [1], affecting about 20% of people worldwide. Approximately 40% of all medical visits are due to chronic pain [2], resulting in annual treatment costs upward of US $600 billion annually in the United States alone [3]. In addition, chronic pain is related to decreased well-being, psychological distress, and high levels of psychiatric comorbidity [4]. Treatment of chronic pain is challenging, and pharmacological approaches are often limited by side effects [5]. Regarding the importance of chronic pain as a major health problem, there is an urgent need for noninvasive, well-tolerated, and easy-to-use alternative treatment options. Besides obvious pharmacological and invasive treatments, there are numerous other physical, psychological, or social approaches that traditionally include psychotherapy, physical therapy, massages, or acupuncture [6,7]. Others have used biofeedback, including heart rate variability–based biofeedback in order to treat chronic pain [8,9]. Another new and promising approach may be the use of virtual reality (VR) [10-12] as a method to enable new forms of treatment.

Based on the findings of an association between chronic pain and abnormal central representation of the body schema and in pain perception [13], several studies showed an association between the experimental manipulation of the body schema using multisensory inputs and pain reduction in patients with chronic pain [14-16]. The term body schema refers to the brain’s continuous representation of the position and movement of the body parts in space by integrating multisensory body related information (eg, visual, tactile, and motor) in order to create a coherent perception of one’s own body [17]. The probably best-known experiment about the manipulation of the body schema using multisensory stimulation is the so-called rubber hand illusion [18]. In the rubber hand illusion experiment, ownership (eg, the degree to which we identify with a given body part) over a fake hand is induced when the participants see it being touched in synchrony with their own physical hand. It is believed that the rubber hand illusion is based on the faulty cross-modal integration of visual (where the participant sees the touch) and sensory (where the participant feels the touch) information into a coherent body schema [18]. Others have highlighted the importance of sensory suggestibility [19] and the characteristics required for the rubber hand illusion to occur [20]. The principle of the rubber hand illusion was successfully extended to the full-body illusion by Lenggenhager and colleagues [21], using VR in order to display tactile stimulation of the participants’ backs synchronously with the visual stimulation of a virtual body. Similar to the rubber hand illusion [14], a change in the perception of pain in healthy participants, as well as in patients with chronic pain, was found in several studies using the full-body illusion [15,22,23], suggesting that the identification with a virtual body induces an analgesic effect for the physical body. However, the currently proposed setup has a major drawback in that it necessitates a second person to perform visuo-tactile stimulation. So in the context of home-based telerehabilitation, there is the urgent need for a method that induces body ownership without the need of another person’s presence.

Recent studies found a way to bypass this problem by using interoceptive biosignals (eg, heartbeat). They were able to induce the full-body illusion, showing the participants a cardiovisual stimulation of a virtual body in VR, with a silhouette flashing around the virtual body in synchrony to the participant’s heartbeat (so-called cardiovisual full body illusion) while participants were not able to (consciously) relate the flashing of the silhouette to the heartbeat [24,25]. Cardiovisual changes influencing self-identification have been linked to altered neural activity in specific brain regions, including the bilateral operculum and parietal somatosensory cortex [25,26]. Moreover, transient changes in neural responses to heartbeats, known as heartbeat-evoked potentials, have been observed in the posterior cingulate cortex and insula during alterations in bodily self-consciousness induced by the full-body illusion [27].

We developed a setup based on the cardiovisual full-body illusion that later could be implemented in the treatment of patients with chronic pain affecting the whole body (eg, fibromyalgia), and not just a body part (eg, complex regional pain syndrome and neuropathic pain after spinal cord injury [15,16]).

In this pilot study, our primary goal was to evaluate an easy-to-use, telerehabilitation-friendly, mobile setup, that reduces the setup complexity significantly by using a stand-alone head-mounted display (HMD) for the visualization and a smartwatch to detect the heartbeats using photoplethysmography (PPG). We hypothesized that participants report a high level of usability and a low number of side effects, allowing for the setup to be tested in a larger randomized clinical trial.

As a secondary outcome, the level of ownership with the virtual body and pain perception were assessed using a randomized within-subject design. We hypothesized that synchronous cardiovisual stimulation results in higher ownership ratings of the virtual body and lower pain perception as compared with asynchronous cardiovisual stimulation.

A total of 20 healthy volunteers were recruited at the University of Bern by word of mouth between April 2021 and January 2022. Inclusion criteria were volunteers aged ≥18 years old, those capable of judgment, those willing to participate in the study (by signing the informed consent form), and those able to follow the study protocol. The exclusion criteria of nonclinical participants were based on self-declaration: no presence of psychosis or major depression with suicidal risk; no history of alcohol or drug abuse; no inability to follow the procedures of the study, for example, due to language problems; no inability to wear an HMD; and no signs of arrythmia, such as atrial fibrillation. As the primary focus of this pilot study was to assess the usability of the device, we aimed for 20 participants, as this number has been shown to be sufficient for usability testing [28].

The study was conducted in accordance with the latest version of the Declaration of Helsinki [26]. The study protocol, including statistical methods, quality control, and data protection, was approved by the Ethical Committee of the canton of Bern, Switzerland (registration number 2019‐01515). All participants were instructed about the study procedure, and an informed consent form was obtained. This resulted in an anonymized dataset, where all identifiable data are archived separately. Study data were collected and managed using REDCap (Research Electronic Data Capture; Vanderbilt University) electronic data capture tools hosted at ARTORG Center for Biomedical Engineering Research [29]. The participants were not financially compensated for taking part in the study.

The algometric measurement method to detect hyperalgesia was carried out by means of a pressure pain provocation test [33]. For this type of algometry, we used a standardized peg with a clamping force of exactly 10 N at an extension of 5 mm (Type Algopeg, size 78×10 mm, polypropylene and nickel, spring reinforced in Switzerland; Annette Kocher, Inselspital Bern). High numerical rating scale values correspond to high pain intensity [34]. The mean values for the right and left ear lobe and the right and left middle finger were calculated and used for statistical analysis.

The pain pressure detection threshold was measured with a standard electronic algometer (Somedic Type II, size 161×170×30 mm, probe of 1 cm²; Somedic Production AB). The electronic algometer was calibrated following the standard protocol as recommended by the manufacturer [35] and set to deliver. Low thresholds correspond to high pain sensitivity [34]. The mean value for the right and the left middle finger was calculated.

Positive values refer to an increase of pain sensitivity (Algopeg) but a lower pain threshold (Somedic), respectively. The difference of the pre- and postintervention measurement for synchronous and asynchronous stimulation was calculated and averaged over the two repetitions.

We assessed ownership for the virtual body using a 7-item questionnaire adapted from Aspell et al [24]. The participants were be asked to indicate how much they agreed with each item using a 7-point Likert scale ranging from 0 (complete disagreement) to 6 (complete agreement). The questions were, for example, “During the experiment, there were times when…I felt as if the virtual body was my body,” “...it felt as if my real body was drifting toward the front,” and “...it seemed as if the flashing semi-transparent template was my heartbeat.”

All questionnaire data were tested for normality (Kolmogorov-Smirnov test of normality) and, if normal, were analyzed using a 1-sided (1-tailed) t test for dependent means or with the Wilcoxon signed rank test. As the quality metric for the algorithm, the root mean square error between corresponding time intervals was calculated over the length of the recordings and over all participants. The significance level was at P<.05. Data were analyzed using RStudio (Posit PBC).

In total, 20 participants (9 male and 11 female) took part in the study. The age ranged from 22 to 29 years. None of them reported prior VR experience.

All the volunteers assessed for the study were eligible. All participants received the intended treatment and finished the study. The data of all participants were included in the analysis.

The results of the SUS were normally distributed and show very high usability and acceptance. The mean value of 4.424 (SD 0.56) was significantly higher than the midpoint of the scale (P<.001; median 4, IQR 4-5). The results regarding side effects measured using the SSQ show minimal to no adverse effects (mean 1.175, SD 0.46; P<.001; median 1, IQR 1-1). For details, please refer to Table 1.

For the validation of the custom peak detection algorithm, its output, meaning the raw list of heartbeat intervals, was compared with the corresponding RR-peak distances (=time) of the ECG reference. Root mean square error between the measured and reference peak-to-peak intervals over all recordings and all participants resulting in 3.4% (SD 2.9%) for the wristwatch and 2% (SD 1.5%) for the watch at the upper arm. This means that the detection algorithm was off in 3.4% and 2% of the cases, respectively. As a second quality measure, we counted the numbers of missed beats of our custom algorithm compared with the ECG reference. Over all participants, more than 6500 beats were detected. The watch located at the upper arm missed only 1 beat, whereas the watch located at the wrist missed 8.

The results from both Algopeg and Somedic were normally distributed.

Ratings for ownership (data were not normally distributed) were both low during the synchronous condition and the asynchronous condition (median in both conditions 0). No significant difference was found between the synchronous and asynchronous conditions (Wilcoxon signed rank test Z=1.007, P=.59).

Our primary goal was to evaluate the reliability of an easy-to-use and mobile VR setup based on the cardiovisual full-body illusion that later could be implemented in the treatment of patients enduring chronic pain. The evaluation included the investigation of the usability and side effects of the new setup as well as the validation of the reliability of the used algorithm. Second, we wanted to test whether synchronous cardiovisual stimulation results in a systematic change of ownership and pain perception in healthy participants. The main finding of this study was that while the VR setup demonstrated high usability and minimal side effects, it did not significantly affect ownership or pain perception.

Regarding the reliability of the used algorithm, we show that the detection of the heartbeat is very reliable and that no relevant deviation of the interpulse-interval compared with the RR interval of ECG was found. Due to its configuration with the sharp edges of the QRS complex, the beat detection in ECG is rather simple [36]. However, the detection of the peak of the blood pulse wave with the same reliability is particularly complicated when live detection, as in the described setup, is required. Due to the position of the PPG measuring sensor being located further apart from the heart (arm and wrist) and wireless communication being slower, compared with the earlier setup using a stationary ECG [24], an additional delay between measuring the peak and displaying it in the HMD is introduced. The ECG measures electric activity and thereby detects the beat even before it reaches the baroreceptors in the periphery. It thus gains some time that can be used for computational time-consuming signal analysis and processing and still show the heartbeat in synchrony with the pulse wave in the periphery. In the new mobile setup, the blood pulse waves in the periphery are used. Thus, less time is available for processing to keep the synchrony between the felt heartbeat and the displayed avatar reaction. The delay introduced by wireless communication and peak detection is around 50 ms, meaning there is a small and constant lag between the pulse wave and the shown beat in the HMD.

In line with the first aim, we found very high usability and no relevant side effects (eg, sickness, oculomotor problems, and disorientation). The results are comparable with the findings of Gerber and colleagues [37]. The presented results are an important prerequisite for the future use of the implemented VR setup in patients even without the help of an experimenter (eg, by patients with a chronic pain condition). Thus, we evaluated the tool to be a reliable, easy-to-use VR setup for the cardiovisual full-body illusion, showing a high level of usability and good tolerance in healthy participants.

The second goal of our study was to investigate the effect of the cardiovisual full-body illusion on ownership and pain perception. We were not able to demonstrate any effect of synchronous cardiovisual stimulation on ownership of the virtual avatar and pain perception. While the pain pressure detection threshold was measured on a continuous scale, ownership and pain intensity were rated on an ordinal scale (Likert scale and numeric rating scale). It can be speculated that a continuous scale would have been more sensitive in order to assess changes of ownership and pain intensity. However, we note that the assessment of the pain pressure detection threshold on a continuous scale did not yield statistically significant results.

While some studies using visuo-tactile stimulation showed a reduction of pain perception in healthy participants [23] and patients with chronic pain [14,15], others did not find any effect of synchronous visuo-tactile stimulation on pain threshold during the rubber hand illusion [38]. In a recent review, Matamala-Gomez and colleagues [12] pointed out that this might be due to the different setups used in the studies. They propose using immersive VR for embodiment of a virtual body, in order to modulate body representation and change pain perception in healthy and clinical populations. Moreover, a previous study has shown that the effectiveness of bodily illusions for pain reduction depends on the visual perspective, for example, the use of a first-person perspective on the virtual body (not a third-person perspective, as in our setup), in order to elicit an analgesic effect [39]. Thus, the absence of ownership and lack of modulation of pain perception in our study might also be related to the use of the third-person perspective.

Also, Solcà et al [16] found that synchronous cardiovisual stimulation reduced pain ratings and improved motor limb function in patients with a chronic regional pain syndrome but not in healthy participants. Thus, this setup needs to be tested further in patients with chronic pain disorders.

Previous studies investigating the effect of cardiovisual stimulation in healthy participants used a video-based setup [24,25]. While we did not directly compare a video-based setup with the computer-generated VR setup using a generic 3D avatar in this study, we speculate that the computer-based VR setup has significant drawbacks if it comes to visual fidelity and visual realism (eg, whether the virtual body resembles the body of the participant). It has been shown that top-down processes, such as visual identity, might influence the rubber hand illusion [40] and the full-body illusion or related paradigms. This is in line with a previous study comparing a video-generated and computer-generated VR setup [41] where no effect on self-location and illusory experience of ownership of more than one body was found in the latter.

First, the sample size of our study was rather small and only included healthy and young participants. This also did not allow us to perform a subgroup analysis. Therefore, it would be interesting to test an improved setup in a larger group of older patients with chronic pain within a randomized clinical trial.

Second, the lack of visual fidelity and realism in the avatar might have influenced the results. Although previous studies suggested that the realism is less important, we cannot comment on this as we did not test this in our study.

Third, the use of a third-person perspective rather than a first-person perspective may have affected the effectiveness of the bodily illusion. Attempts to mitigate these limitations included optimizing the filtering and peak detection algorithms to improve measurement precision.

Fourth, our study was limited to a laboratory setting, and the results may not directly translate to real-world clinical environments.

Finally, it would have been interesting to measure the perceptual drift, for example, if ownership over a virtual body results in dislocation of the perceived self-location from the physical body to the virtual body [21,24,42]. However, given that our study was not able to induce ownership with the virtual body, we do not think that measuring the perceptual drift would have been meaningful in this pilot study. Also, participants already had to be tested for pain threshold and sensitivity as well as ownership right after the manipulation. Therefore, an additional drift measurement seemed not to be feasible or a priority to us at the conception of the study.

In conclusion, we were able to evaluate a new mobile VR setup that theoretically would allow for the treatment of patients with chronic pain in a home-based telerehabilitation setting. While the setup shows high usability and reliability and seems to be well tolerated, no significant effect regarding the level of ownership with the virtual avatar and modulation of pain perception in healthy young participants could be demonstrated in this study.

Future research should focus on improving the visual fidelity and realism of the virtual avatar and testing the setup with a first-person perspective. In addition, larger and more diverse samples, including older patients with chronic pain, should be included to better understand the potential of this VR intervention.