Training of Relaxation Breathing in Virtual Reality Using Biofeedback: Comparison Study

This study evaluates the efficacy of a virtual reality (VR)-based relaxation intervention incorporating real-time respiratory biofeedback, delivered via a standalone Meta Quest 2 headset. This study uses a randomized crossover design in which all participants undergo two sessions of virtual reality (VR) relaxation with biofeedback, followed by two sessions without biofeedback, or vice versa. The order of conditions is randomized to control for order effects. The randomization is done via an online randomization tool.

The intervention utilizes a custom-developed application-BreezeTerraVR-designed by the research team at the Center for Virtual Reality Research in Mental Health and Neuroscience, part of the National Institute of Mental Health (NIMH), Czech Republic. BreezeTerraVR is a structured and modifiable VR application specifically built to facilitate training in slow-paced, rhythmic breathing, allowing flexible adjustments to the inhalation, exhalation, and breath-hold phases according to research needs or participant abilities.

The virtual environment represents a calming natural meadow, surrounded by trees. Participants are immersed in this nature scene and guided to practice square (box) breathing, consisting of 4 seconds inhale, 4 seconds hold, 4 seconds exhale, and 4 seconds hold. The target breathing rhythm is visualized through a colored mist moving between two concentric mushroom-shaped circles in the visual field.

Two distinct biofeedback modalities are integrated into the VR experience for the intervention condition with BF:

• Real-time biofeedback is represented by an animated deer walking in front of the user. The deer's exhalation is visualized with visible steam from its nostrils, mirroring the participant's actual breathing rhythm as measured by a Go Direct® Respiration Belt (Vernier).
• Targeted biofeedback is represented by a tree located at the center of the mushroom circle. This tree dynamically grows or shrinks depending on how closely the participant's breathing matches the set rhythm. The feedback difficulty can be adjusted to four levels (0-3), the study protocol applied Level 2 (moderate): Accurate rhythm required for growth; the tree can shrink if the breathing is out of a given rhythm.

In the control condition (Intervention 2), participants engage with the same VR environment and breathing pattern but without biofeedback elements: the deer and the tree are not present. Visual and auditory guidance is still provided via the mist and voice instructions, but there is no adaptive feedback.

All recruited participants sign an Informed consent to participate in the study.

The study duration is three weeks with 4 relaxation sessions in total. The participants complete two VR relaxations per week (e.g.Tuesdays and Thursdays). After the first week of interventions, a one-week break is implemented. The third week the same protocol with two additional sessions is completed. This means the participants complete one of the interventions (conditions) in the first week and the other intervention in the third week.

Each study session takes no longer than 25 minutes per participant. Each of the 4 sessions includes a pre-intervention baseline assessment, a standardized VR breathing session (5 minutes duration) with 3 additional minutes before and 3 minutes after the relaxation spent in the VR environment to measure the pre/post respiratory rate, and a post-intervention assessment. The participants complete the given questionnaires on tablets or computers.

The primary outcomes measured before and after each session include:

• Subjective change in anxiety
• Self-reported relaxation state
• Objective physiological change in respiratory rate (measured for 3 minutes before and after the relaxation).

Secondary outcomes completed only once, immediately after the final Session 4, include the sense of presence and cybersickness measures.

Data will be analyzed using appropriate statistical methods based on sample size and distribution characteristics. Parametric or non-parametric tests (e.g., t-tests, ANOVA, LLM, Mann-Whitney U, or Wilcoxon signed-rank tests) will be used to evaluate within- and between-group differences in anxiety reduction, relaxation level, and physiological measures. Additional analyses may explore interactions between biofeedback accuracy, presence, and outcome effectiveness.

Inclusion criteria

• patients with Anxiety disorders

Exclusion criteria

• severe neurological disease (e.g. epilepsia)
• psychiatric disorders other then Anxiety disorder
• serious somatic disease