Ethics approval for this study was provided by the Ochsner Health Institutional Review Board (2024.054). Informed consent was waived for clinician participants. All interview data collected for this study were fully de-identified and anonymized prior to analysis to ensure participant privacy and confidentiality. Secure storage protocols were implemented in compliance with institutional and federal data protection standards. Access to the data was restricted to authorized study personnel directly involved in data collection and analysis. No compensation was provided to study participants. All procedures were conducted in accordance with appropriate ethical guidelines and applicable regulations. All data and materials generated in this study were managed in accordance with the FAIR (Findable, Accessible, Interoperable, and Reusable) data principles, ensuring that de-identified qualitative transcripts, analysis, and related materials can be made available upon reasonable request to promote transparency and reproducibility.

An HMD-based VR optokinetic drum was developed using a human-centered design approach as described in the VR-CORE Model (31). This paper reports results from the VR1 stage, the first of the three phases of clinical design, focusing on eliciting feedback from end-user neurorehabilitation clinicians. The OVE was developed by a multidisciplinary team including a neurorehabilitation physician (KS), neuro-physical therapist (KD), an MXR developer (AS), and a biopsychologist (NVP).

The Academy of Neurologic Physical Therapy of the American Physical Therapy Association’s recent update to clinical practice guidelines for vestibular rehabilitation recommends habitation exercises whereby individuals are systematically exposed to provocative stimuli that, with repeated exposures, reduces symptomology (23). Optokinetic stimuli are specifically recommended as a tool for habituation and visual motion desensitization. Additionally, optokinetic stimuli challenges the visual system therefore improving integration with vestibular and somatosensory inputs. Finally, optokinetic stimuli can enable contextual adaptation to visually complex and realistic environment. This approach supports improved tolerance of motion-rich setting encountered in daily life and informed the development of the OVE prototype (Figure 2).

Unity, a real-time 3D engine, served as the platform for development, facilitating the creation of an immersive virtual environment. GitHub, a code-hosting platform, aided in development and version control. The MXR developer designed and created dynamic VR optokinetic stimulus tailored for neurorehabilitation. A Meta Quest 3 HMD was used to display the optokinetic environment.

The study recruited 10 neurorehabilitation clinicians from 5 different clinics with experience and interest in vestibular rehabilitation. The sample size is consistent with the design of similar studies and the general recommendations for qualitative analysis (36–38). To be included in the study, clinicians were required to be physical or occupational therapists with > 2 years of experience treating patients with vestibular disease. Individuals who were unable to tolerate the HMD or OVE exposure were excluded. Ten clinicians met criteria and were offered a chance to participate. One clinician did not complete the study due to lack of clinical experience.

Participants were initially educated on the HMD, including how to properly don headset and utilize controllers. After initial set up of HMD, study lead accessed the OVE on HMD, to allow the clinician to view application. Clinicians viewed optokinetic background with both pass through and fully immersive setting (Figure 4). Using the controllers, clinicians navigated menu features to adjust visual parameters including speed of lines (scale of 1–10), number of dots (1 or 2), distance between dots, dot depth, and speed of dots (scale of 1–10). Experience with the OVE was self-guided, and while time was not formally collected, participants spent between 5 and 15 min engaging with the application. HMDs were manually cleaned using disinfectant wipes between uses.

Semi-structured interviews were conducted with neurorehabilitation clinicians between October 2024 and March 2025 (Supplementary Table 1). All interviews were led by a neuro-physical therapy expert (KD). Interviews were recorded and transcribed. Discussions lasted approximately 30 min and included open-ended questions as well as pre-defined probes. The interview guide focused on three main areas: 1. Usability of the HMD-based application, 2. Approaches to improve the utility and experience of the OVE, and 3. Methods and barriers to integrating into existing clinical workflows.

A word cloud visualization (Figure 3) was generated using RStudio to display all words appearing more than once, with word size proportional to frequency and color indicating relative frequency across the data. Word-level preprocessing included conversion to lowercase, removal of punctuation (including dashes and quotation marks), common stop-words (e.g., “also,” “just,” “like,” “can”), and filtering out unique-instance terms.

Coding and theme identification were conducted using a reflexive thematic analysis approach as described by Braun and Clarke (39). Initial familiarization and open coding were independently performed by a physician-scientist with experience in neurorehabilitation and qualitative research (KS) and a neuro-physical therapist with vestibular expertise (KD) and a third team member (SS) reviewed coded transcripts to support consistency and completeness. Coding disagreements were addressed through iterative discussion until consensus was reached, with codes refined or merged as appropriate. This consensus-based approach emphasized shared interpretation rather than inter-rater reliability metrics, consistent with reflexive thematic analysis methodology. To support reflexivity, the analytic team met bi-monthly to discuss how professional backgrounds influence interpretation and generalizability. Notes were maintained during coding to document analytic decisions and evolving theme definitions. Final themes were defined, named, and reviewed collaboratively, and all themes were supported by representative participant quotations included in the Results and summary tables.

Thematic saturation was assessed iteratively during data collection and analysis. Saturation was considered achieved when successive interviews failed to generate novel codes or meaningfully alter existing themes. In this study, saturation was reached after eight interviews; two additional interviews were conducted and confirmed thematic stability, yielding no new codes or themes. To further illustrate coding outcomes, we have added representative excerpts and consolidated participant quotations within thematic tables, enabling readers to directly link raw data to analytic conclusions.

Generative AI (ChatGPT) was used to (1) assist with literature synthesis, (2) support organization and summarization of already-identified codes and themes, and (3) improve clarity and consistency of manuscript language. Importantly, AI tools were not used to generate primary codes, define themes independently, interpret raw interview transcripts, or make analytic decisions. All coding, theme development, and interpretive judgments were performed by the human research team. Outputs generated with AI assistance were reviewed, edited, and validated by the authors to ensure fidelity to the original data and to maintain analytic integrity.

Clinician participants expressed positive perceptions about using a medical extended reality application as an adjunct to vestibular therapy. Clinicians rated the likelihood of using the OVE in clinical practice as high, with Likert scores ranging from 6 to 10 (median = 9) with 10 representing highly likely. The primary potential limitation cited was variable patient tolerance. Of note, no clinician participants noted adverse effects, like vertigo or nausea, related to the use of the VR HMD (Figure 4, top).

The analysis revealed five overarching themes (Figure 5): (1) clinical usability and setup, (2) control and customization needs, (3) immersive design and realism, (4) output and measurement preferences, and (5) implementation barriers.

There was general enthusiasm across participants to utilize VR in therapy consistent with literature on its use in rehabilitation (41). There, however, remains a general hesitancy of using VR due to concerns about whether this technology can effectively be operationalized in a busy clinic with minimal setup burden and minimal interruption to therapeutic flow. This emphasis is consistent with prior clinical VR literature describing how HMDs can introduce friction in actual clinical implementations through hardware management, required technical proficiency, and mid-session adjustments that disrupt therapy (42).

Accordingly, clinicians in our sample prioritized therapist-directed control (ideally externalized) to adjust stimulus parameters without removing the headset. This preference is also congruent with broader human factors considerations in rehabilitation technology: tools that preserve clinician agency and reduce session disruption are more likely to be adopted than those requiring frequent device interaction by the patient or clinician during care delivery (43).

Another key clinical concern raised by participants was patient usability and tolerance, particularly among individuals who already suffer from motion sensitivity. This concern is well supported by the cybersickness literature, where VR is associated with disorientation, nausea, and oculomotor symptoms in susceptible users, and where physiologic perturbations have been observed during cybersickness episodes (44).

Importantly, “exercise performance” in vestibular contexts is often constrained by symptom provocation thresholds rather than strength or endurance limitations alone. Thus, device factors like fit, weight, heat, sensory load, and the ability to grade exposure are particularly relevant in the use of VR in those with vestibular dysfunction. Seo et al. demonstrated that overall usability of VR HMDs will significantly impact whether a VR intervention can be delivered effectively. In parallel, the broader HMD balance literature suggests that head-mounted VR systems are increasingly capable for balance-related assessment and training (45), but that outcomes vary substantially by population, protocol, experiential complexity and device parameters which reinforces the need for staged validation that goes beyond early feasibility and focuses on usability (46–48).

Clinicians consistently requested session summaries and objective metrics like tolerance time, head movement speed, and stimulus velocity. This feedback reflects a need to translate a VR experience into the language of clinical evaluation and management. This aligns with a broader movement in rehabilitation toward instrumented feedback, where sensor-enabled systems support auditing of dose, adherence, and real-world activity patterns. For example, the SIRRACT trial demonstrated that wireless sensing could provide “ground truth” about patient activity patterns in inpatient stroke rehabilitation (49). Even though this additional feedback did not always translate into improved outcomes, it highlights that measurement infrastructure can still be clinically valuable for visibility and workflow integration. In vestibular rehabilitation, clinician-facing dashboards can serve as a practical bridge between immersive stimuli and standard clinical workflows by enabling rapid parameter adjustments, documentation support, and longitudinal progress tracking.

Participants’ interest in quantification and biofeedback is consistent with the wearable sensor literature describing the clinical utility of Inertial Measure Unit (IMUs). IMUs are devices that measure and report body specific forces, angular rates, and orientation. IMUs capture movement kinematics and support feedback-enabled rehabilitation paradigms (50). In vestibular-specific applications, IMU-guided systems have also been explored for real-time feedback on head rotation angles and speeds. IMUs have been able to determine peak head turning speeds in patients after vestibular schwannoma resection, with improvements in speed following vestibular rehabilitation (51). Additionally, IMUs have been used during positional maneuvers, supporting the concept that IMU-derived metrics can meaningfully augment vestibular care delivery when appropriately designed (52). Together, these data support the rationale for incorporating IMU-driven session summaries and therapist dashboards in future iterations of the OVE to better align with clinician expectations for actionable metrics that support patient-specific treatments.

Beyond usability and tolerance, clinicians raised pragmatic concerns ranging from cleaning protocols, device storage/charging, and storage/space constraints. These issues are increasingly recognized as real adoption determinants for MXR. Recent literature has noted the absence and need for a standardized approach to disinfecting VR headsets (53). Addressing these basic logistical issues is likely necessary for successful real-world clinical deployment.

This study is limited in generalizability due to a small sample size from a single institution that only included clinicians. Thematic saturation suggests, however, that a larger sample size would likely only yield incremental value. Inclusion of therapists from different sites or organizations would support external validation. Clinician participants were non-randomized, unblinded colleagues who may have influenced each other’s responses and introduced a diffusion of treatment effect as described by Urban (54).

Participants’ feedback emphasized the need for streamlined controls, increased environmental variation including pass-through and realistic environments, enhanced quantification of sessions, and approaches to optimize clinical integration. Future iterations of the OVE will include a tablet-based interface focused on improving the clinician and patient experience (Figure 6). Future versions of the OVE will incorporate network code to support streaming of the HMD view to a tablet so that clinicians can observe the patient experience in real-time. The interface will also include graphic shaders allowing clinician-driven and dynamic adjustment of stimulus parameters such as speed, width, color, direction, and pattern type within a patient session. Clinicians will have the ability to save and restore stimulus settings to enable efficient clinical utilization and integration.

To support real-world relevance and clinical generalizability, the visual environments will be expanded to include various abstract optokinetic stimuli that can be implemented with either complete immersion or pass-through, allowing for a blend of digital and physical realities. Future versions will also include immersive settings that mimic common dizziness triggers, such as grocery stores, cafes, and driving (Figure 7). Clinician-facing dashboards will be incorporated to provide session summaries and basic patient performance metrics such as time-used and repetitions performed. Future iterations may explore integration of eye-tracking and wearable sensors to support exploratory measurement of physiological and behavioral signals in subsequent usability and efficacy studies. Additionally, implementation barriers like clinical space needs, charging, and cleaning HMDs will be addressed.

Future studies will involve collecting feedback from a larger and more diverse cohort of clinicians as well as vestibular therapy patients. This will inform a finalized OVE application whose safety, tolerance, and efficacy will be evaluated in a randomized controlled clinical trial as consistent with VR2 and VR3 of the VR-CORE framework. Spending adequate time and effort in VR1 to address clinician and patient experience feedback is necessary prior to progressing to VR2 investigations. In addition, addressing implementation barriers will optimize future implementation studies of VR-supported vestibular rehabilitation across diverse care settings, including under-resourced clinics and remote rehabilitation scenarios.