Developing immersive virtual exposures for obsessive-compulsive disorder – Protocol for a Randomised, Controlled, Proof-of-Concept Feasibility Trial

Developing immersive virtual exposures for obsessive-compulsive disorder – Protocol for a Randomised, Controlled, Proof-of-Concept Feasibility Trial | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Method Article Developing immersive virtual exposures for obsessive-compulsive disorder – Protocol for a Randomised, Controlled, Proof-of-Concept Feasibility Trial Mariana Pinto da Costa, André Abreu, Carolina Fialho, Maximin Lange, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9244535/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Exposure and response prevention (ERP) is the gold-standard psychological treatment for obsessive compulsive disorder (OCD). However, its delivery typically requires frequent therapist availability and repeated patient travel to treatment settings, which limits accessibility. In addition, the idiosyncratic nature of obsessive-compulsive symptoms presents challenges for conducting effective exposures within the time and material constraints of traditional clinical environments. Virtual reality (VR)-based interventions may help address these limitations. This study aims to assess the feasibility of a novel approach that uses generative artificial intelligence (GenAI) to create personalised, immersive 3D exposure environments tailored to individual patient fears. Methods This is a randomised controlled feasibility study with three parallel arms and an assessor-blinded design. Forty-five adults with a primary diagnosis of OCD and moderate to extremely severe symptoms will be randomly allocated in a 1:1:1 ratio to OCD-related exposure administered via VR environments, neutral VR environments, or OCD-related stimuli presented on a widescreen display. Participants will complete a baseline assessment and an GenAI-based stimulus titration session, followed by two therapist-led ERP sessions that frame five consecutive days of asynchronous exposure (exposure blocks/scenarios that are not therapist-led ERP). The intervention uses text-to-image synthesis, image conversion into 3D Gaussian Splatting environments, and delivery via VR headsets or widescreen display. Primary feasibility outcomes include recruitment and retention rates, data completeness, and adherence to the asynchronous exposure protocol. Secondary outcomes include progression through personalised exposure hierarchies, physiological reactivity (electrodermal activity and heart rate), cybersickness, and subjective distress measures. Discussion This study will determine whether AI-driven VR exposure is feasible, safe, and acceptable for adults with OCD. The proposed approach standardises the process of content generation while personalising the actual stimuli, towards leveraging technology to ensure the personalised needs of these patients are more effectively met. Results can inform the refinement of the intervention and the study procedures, including sample size estimation for a future randomised controlled trial. If successful, this methodology could have the potential to improve scalability, reduce costs, and enhance the ecological validity of exposure therapy while maintaining clinical efficacy. Trial registration ISRCTN13869986. Registered 29 December 2025. Prospectively registered. Obsessive compulsive disorder Virtual reality Exposure and response prevention Psychotherapy Generative artificial intelligence Feasibility study Personalised treatment 3D Gaussian Splatting Immersive technology Figures Figure 1 Figure 2 Introduction Cognitive-behavioural therapy (CBT), particularly exposure and response prevention (ERP), is the gold-standard treatment for obsessive-compulsive disorder (OCD) (1,2). ERP involves the gradual exposure to anxiety-provoking stimuli while refraining from compulsive behaviours intended to reduce distress. Exposure can be conducted in vivo, imaginally, or interoceptively (3–5). Despite its efficacy, ERP is associated with several challenges that can limit its accessibility and uptake., While recent reviews report treatment dropout rates of approximately 10-16% (7–9), earlier reviews reported rates of up to 25-30% (6) . Attrition has been attributed to the high emotional burden experienced by patients, particularly during early treatment phases, when the effort required to confront anxiety-provoking stimuli without performing their compulsive soothing rituals may seem harder to endure (10,11). ERP typically requires regular therapist contact, with frequent sessions and adherence to homework exposure tasks, commonly leading to scheduling difficulties and impose additional time and travel costs for patients (8,12,13). The clinical setting also imposes inherent constraints on treatment delivery, as therapists often lack the means to replicate the idiosyncratic, context-bound triggers for specific obsessions within the session confines, which risks diminishing ecological validity of the proposed exposure exercises and shifts the burden to the patient’s self-motivation and compliance (14). Generative Artificial Intelligence (GenAI) and virtual reality (VR) technologies may offer opportunities to address these barriers. These tools can facilitate personalised, immersive therapeutic interventions that may be delivered at scale and reduced cost (15,16). GenAI models can simulate OCD-related stimuli in controlled environments, supporting the development of tailored exposure scenarios (17). Recent meta-analyses of virtual reality exposure therapy in anxiety and obsessive-compulsive spectrum disorders have reported mixed results (18–20). A key limitation of existing VR interventions for OCD is their reliance on pre-built, standardised environments (e.g., dirty bathrooms, cluttered kitchens, or disorderly rooms) which are often difficult to adapt to the heterogeneous and idiosyncratic symptom presentations characteristic of the disorder. This is problematic given that VR exposures only elicit therapeutic responses when the stimuli closely match the individual’s fears and compulsions (21). This highlights the need to better integrate the idiosyncratic nature of effective ERP into a standardised delivery model, with individualised exposure content, ensuring that protocol rigidity does not come at the expense of ecological validity. To address this need, a novel approach standardising the process of creating exposure scenarios while personalising the scenarios to individual patients was developed. This methodology, using AI-driven stimulus creation with structured titration, allows protocol replication across sites, while maintaining the individualisation necessary for effective exposure therapy. This approach introduces several methodological and technical innovations over previous OCD VR and mixed reality research. It combines the rapid, on-demand generation of personalised, photorealistic 3D exposure environments using GenAI and Gaussian Splatting, thereby eliminating the need for pre-built assets, 3D modelling, or game engine development. However, it remains unclear whether AI-generated, individualised VR exposures are feasible and acceptable in clinical practice. Thus, this article reports the protocol of the DIVE-OCD ( Developing Immersive Virtual Exposures for OCD) randomised controlled feasibility trial. Objectives The primary objective of this study is to evaluate the feasibility, acceptability, and safety of using generative artificial intelligence to produce individualised virtual reality exposure stimuli for patients with OCD. Secondary objectives are to collect data on the impact of the intervention in the physiological reactivity, cybersickness, general distress, and exposure distress. This study will explore emotional engagement with the exposure content when delivered via immersive virtual reality (VR) and via a screen-based format, as well as in comparison to neutral content. Methods Patient and public involvement Three patient and public involvement (PPI) members, including two people with lived experience of OCD and one family member, contribute to the design, interpretation of findings and dissemination. During the design phase, PPI members review study materials to ensure language clarity and accessibility. During the analytic phase, they will review and help refine the wording of themes. During the reporting phase, they will assist in co-creating lay summaries, advise on dissemination strategies, and help communicate findings to local organisations and other individuals with lived experience. Two of these members also serve on the board of a Portuguese OCD association ( POC Doentes e Famílias), supporting the dissemination of evidence-based treatment options and resources for people with OCD. Trial design This is a randomised controlled feasibility study with three parallel arms and single-blind (assessor-blinded) design, stratified by current therapy status (non-ERP versus ERP-ongoing). Participants are allocated in a 1:1:1 ratio to receive exposure via OCD-related VR environments, neutral VR environments, or OCD-related stimuli presented on a screen. It employs an exploratory framework to evaluate the feasibility, acceptability and safety of GenAI-driven virtual reality technology as a tool to enhance adherence to exposure and response prevention therapy in OCD treatment. This protocol follows the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guideline (22) (see Additional file 1 for the SPIRIT checklist). Trial setting Participants are recruited at the Champalimaud Clinical Centre (CCC), Lisbon, Portugal, a private hospital with a neuropsychiatry unit, and integrated research programs. All intervention sessions are delivered on-site. Participants’ eligibility criteria Inclusion criteria: Adults aged 18 to 65 years Primary diagnosis of OCD (ICD-10 F42.0, F42.1, F42.2) confirmed by Mini International Neuropsychiatric Interview (MINI 5.0.0) Y-BOCS-II score ≥ 14 (moderate to extremely severe symptoms) At least one OCD subtype amenable to visual exposure (e.g. contamination/washing, checking, or symmetry/ordering) Ability to provide informed consent Exclusion criteria: Predominant OCD symptoms not suitable for visual exposure Current or past psychotic episode, schizophrenia, or autism spectrum disorder Current substance use disorder Self-reported history of severe cardiovascular disease where acute physiological arousal poses a safety risk, defined as the presence of an implanted cardiac device (pacemaker or ICD), a diagnosis of heart failure or aortic aneurysm, or a major cardiac event (myocardial infarction, stroke, or cardiac surgery) within the previous 12 months Visual or auditory impairment, susceptibility to motion sickness, or epilepsy Eligibility criteria for those delivering interventions Study Lead Clinician (unblinded): clinical psychologist with more than two years of clinical experience and training in ERP for OCD; conducts baseline assessments, AI stimuli titration after training from machine learning experts, and supervises asynchronous exposure sessions (i.e. supervised exposure blocks/scenarios that are not therapist-led ERP). Study Support Clinician (SSC) (unblinded): psychologists or psychiatrists, who assist the SLC with recruitment, eligibility assessments, and monitoring asynchronous exposure sessions. SSC who are psychiatrists, provide immediate consultation for any urgent psychiatric care urgent needs arising during the study. ERP Therapist (blinded): delivers the psychotherapeutic sessions (Visit 2A and Visit 7A) after receiving specific training on the manualised intervention. Informed consent Written informed consent is obtained by the research team following review of the participant information sheet and discussion of study procedures. Intervention and comparators descriptions The trial includes three arms (Figure 1): OCD-VR arm: immersive personalised GenAI-generated VR environments corresponding to the participant’s specific OCD symptoms. Neutral-VR arm: immersive personalised GenAI-generated VR environments described by participants as non-threatening, e.g., natural landscapes. OCD-Screen arm: the same personalised GenAI-generated OCD-related stimuli as in the OCD-VR arm, delivered via a widescreen display. Comparing AI-generated stimuli in VR versus widescreen isolates immersion effects, while OCD-VR versus Neutral-VR tests content-specific effects. The intervention follows a sequential structure combining standard psychotherapeutic care with the experimental exposure component. Figure 2 displays the study components in both experimental and control conditions. The protocol is delivered over seven visits. Table 1 displays an overview of these visits. Table 1. Overview of study visits and experimental procedures. Visit 1: Assessment and Titration Participants attend an initial session for diagnostic confirmation and stimulus generation. The Study Lead Clinician (SLC) or Study Support Clinician (SSC) conducts informed consent, confirms the OCD diagnosis and screens for possible diagnostic exclusion criteria using the Mini International Neuropsychiatric Interview (MINI 5.0). The SLC then assesses OCD severity and explores symptom description with the Y-BOCS-II, provides psychoeducation, and collaboratively constructs a personalised symptom hierarchy, ensuring all items fall within clinically appropriate intensity levels (SUDS ≤ 8). During this session, participants also complete the AI-stimuli titration. They describe their triggers to generate sample images, which are iteratively refined with the SLC until clinically valid and within the target SUDS range. Individualised 2D images are generated using the FLUX.1 [dev] text-to-image model (Black Forest Labs, 2024) via the fal-ai/flux/dev endpoint using the fal-client Python library (v0.9.1 or higher) to ensure reproducibility. Prompt engineering prioritises detailed sensory descriptors over abstract concepts to ensure high ecological validity. All images undergo dual-layer safety review: automated model-based filters and manual inspection by the SLC to ensure appropriateness and therapeutic relevance. Images are only shown to participants after explicit approval by the SLC. Three categories of OCD-related stimuli and three categories of neutral images are finalised per participant. All generation parameters (prompt text, seed, model version, timestamps) are stored as JSON files, and all tasks are performed on a Lenovo LOQ Gaming 17i workstation to ensure consistent performance. 3D Environment Generation Approved 2D stimuli are used to generate immersive 3D environments using the Marble platform developed by World Labs (2024). Marble produces navigable volumetric scenes from image inputs using an GenAI-based 3D scene generation pipeline that represents environments with 3D Gaussian Splatting (3DGS). The resulting scenes preserve fine visual detail and allow real-time viewpoint changes. The generated scenes are exported and rendered using SparkJS for interactive viewing. VR exposures are delivered on Meta Quest 3 headsets with calibration for interpupillary distance and starting position. All scenes are pre-inspected by the Study Lead Clinician (SLC) to confirm visual quality and clinical suitability. Screen-based exposures are delivered via an 85-inch Neo QLED 4K display. The same 3DGS environments are used in both VR and screen conditions to ensure identical visual stimuli across modalities. The difference between modalities may contribute to outcomes. VR participants view the scene with natural 6DoF head tracking, while screen participants use a joystick to control camera orientation and navigation. Visits 2A and 2B: First ERP and First Asynchronous Session Approximately one week after Visit 1, participants attend a 90-minute ERP session (Visit 2A) led by the blinded ERP Therapist. Exposure begins with lower-ranked items in the symptom hierarchy (progression interrupted if SUDS > 8). On the same day, participants complete the first of five daily 30-minute asynchronous sessions (Visit 2B) under supervision of the SLC or SSC. Physiological data (electrodermal activity and heart rate) are collected continuously, with subjective distress measured verbally at scene transitions. Visits 3–6: Asynchronous Phase Participants complete four further 30-minute asynchronous exposure sessions on consecutive working days. Procedures, data collection, and supervision are identical to Visit 2B. Visit 7A: Second ERP Session One day after the final asynchronous phase visit, or at the next available opportunity within a 72-hour window, and approximately one week after Visit 2, participants attend a second 90-minute ERP session to progress further up the SUDS hierarchy under the guidance of the same blinded ERP Therapist. Visit 7B: Qualitative Debriefing Immediately following Visit 7A, participants engage in a structured qualitative interview with the SLC to provide feedback on feasibility, acceptability, and experience of the intervention. Strategies to improve adherence to intervention/comparator: On the first visit, the SLC or SSC explains the study procedures and answers any questions. All sessions and data collection occur in person, providing additional opportunities to encourage engagement. Participants receive daily text and email reminders for the upcoming appointments. If a session is missed, the SLC contacts the participant to reschedule the session and discusses any barriers to attendance. Criteria for discontinuing or modifying allocated intervention/comparator There is no participant relocation. If a participant is unable to continue with asynchronous exposure in VR or screen or experience any adverse events resulting in their wish to interrupt the asynchronous phase, their participation is discontinued from this intervention component. For an intention-to-treat analysis, these participants remain eligible for all subsequent outcome measures; specifically, they are invited to complete the second ERP visit and the qualitative debrief interview, unless data removal is explicitly requested. Concomitant care permitted or prohibited during the trial: Participants must maintain their current medication regimen as prescribed. Regarding psychotherapeutic interventions, participants are divided into two groups: Group A: Participants not currently receiving ERP must agree not to initiate any new ERP treatment during their participation in the study. Group B: Participants currently undergoing ERP, together with their treating therapists, must suspend all external ERP sessions and homework during the five consecutive-day asynchronous exposure phase (between the first and second study ERP sessions). Immediate supportive care is available on-site for adverse events, namely if a participant experiences severe anxiety or motion sickness. Ancillary and post-trial care: All participants are under outpatient psychiatric care in the CCC. Post-trial care is provided by their respective clinicians, who will be informed of any adverse events. Outcomes: The full assessment schedule is shown in Table 2. Primary outcome Feasibility: Recruitment rate : number of participants enrolled within the recruitment window Retention rate : proportion of enrolled participants completing the final assessment Data completeness : proportion of expected data points collected across all measures Adherence : proportion of planned exposure sessions completed in full (15 minutes) Safety: assessed through recording of any adverse events throughout the study period. Acceptability: assessed through end of study qualitative interviews. Secondary outcomes Physiological reactivity: measured by electrodermal activity (microsiemens) and heart rate (beats per minute) standardised within participants relative to within-session baseline. Cybersickness : measured by the Simulator Sickness Questionnaire (SSQ) collected after each VR session (Portuguese version - SSQp; (23)). General distress: measured by the visual analogue scale (0-10) and State-Trait Anxiety Inventory State version (STAI-Y1) collected after each exposure session (Portuguese version; (24)). Subjective distress: Change in the percent hierarchy position score (progression through the individualized SUDS hierarchy) from Session 2A (approximately 7 days post-randomisation) to Session 7A (approximately 14 days post-randomisation). Adverse Events An Adverse Event (AE) is any untoward medical or psychological occurrence, regardless of causal relationship with the study intervention. Given the nature of ERP, where anxiety induction is a therapeutic mechanism, a distinction is maintained between expected therapeutic distress and adverse events. High subjective distress (e.g., SUDS ratings >8/10) during exposure is considered therapeutic; however, distress that persists significantly beyond the session, panic attacks requiring medical support, or severe refusal to eat or drink due to symptom exacerbation are classified as AEs. A Serious Adverse Event (SAE) is defined as any event that results in death, is life-threatening, requires inpatient hospitalisation, results in persistent or significant disability/incapacity, or requires intervention to prevent such outcomes. Potential harms anticipated in this study are consistent with those commonly associated with exposure therapy and VR research. Exposure to feared stimuli may temporarily increase anxiety or discomfort; to mitigate this risk, exposure hierarchies are constructed to exclude stimuli expected to elicit Subjective Units of Distress (SUDS) ratings above 8, and clinicians are present during all sessions to pause or terminate exposure as needed. Cybersickness or visual discomfort may occur in the VR condition and will be routinely monitored using the Simulator Sickness Questionnaire, with the option to interrupt or reschedule sessions if symptoms arise. The use of AI generation can introduce a theoretical risk of generating clinically inappropriate or disturbing content. To minimise this risk, all images will be screened through algorithmic safety guardrails and manually reviewed by the SLC before being shown to participants, ensuring that only clinically appropriate stimuli are used. Data and privacy risks are minimised through strict adherence to GDPR-compliant local storage: all immersive content will run on the VR or widescreen systems, and physiological recordings will be stored securely on institutional servers without cloud transfer. Collectively, these procedures are intended to identify, manage and document adverse events, supporting an assessment of the safety and feasibility of AI-driven immersive exposure within ERP for OCD. Monitoring and management Safety is monitored throughout the study by the unblinded SLC or SSC and the blinded ERP Therapist across the three arms. Specific intervention-related risks are managed via the following protocols: Cybersickness : Visually induced motion sickness is monitored using the Simulator Sickness Questionnaire (SSQ) after each VR session. Acute symptoms (e.g., severe nausea) trigger immediate session cessation. Suicidality : Monitored via clinical observation and the MINI interview. Any expression of suicidal ideation or intent triggers an immediate assessment by the on-site psychiatric team. Psychological distress : Sessions are terminated immediately upon participant request or if unmanageable distress (e.g., dissociation) is observed. Reporting and causality All AEs will be recorded in the Case Report Form (CRF) and the AE log. Severity is graded on a three-tier scale (Mild, Moderate, Severe), and causality is assessed using a five-point scale (Not Related, Unlikely, Possible, Probable, Definitely). SAEs will be reported by the Principal Investigator to the on-site Research Ethics Committee (REC) within 15 days of the study team becoming aware of the event. In the case of urgent safety measures required to protect participants from immediate hazard, the REC will be notified within 3 days. Participant timeline See Table 2 for details on the participant schedule of enrolment, interventions, and assessments. Participants who are interested in taking part in the study attend the CCC for Visit 1, where they sign the informed consent form and complete the screening. If eligible, participants complete psychiatric diagnostic assessments, receive psychoeducation, collaboratively construct a personalised symptom hierarchy, and complete the stimulus preparation task with the SLC. Following this visit, participants are randomised. Participants return the following week for Visit 2A, the first ERP session with a clinical psychologist, followed by Visit 2B the same day, the first asynchronous session supervised by the SLC or SSC. Visits 3-6 occur the following 4 days and are similar asynchronous sessions. Participants then attend Visit 7A, the second ERP session, followed by session 7B, where participants engage in a qualitative interview with the SLC. Table 2. Participant timeline: schedule of enrolment, interventions, and assessments Sample size The target sample size for this feasibility trial is 45 participants, with 15 allocated to each of the three study arms. Following Montgomery (25), a sample of 15 per arm was chosen to prioritise the estimation of key parameters, specifically standard deviations and effect sizes, required to conduct a power analysis for a future RCT. Recruitment Participant identification and referral Proactive and multi-channel recruitment strategies are implemented, including: (1) direct clinical referral from psychiatrists and psychologists within CCC (2) database systematic screening of the CCC electronic health records to identify patients with a primary diagnosis of OCD (ICD-10 F42.0, F42.1, F42.2) who meet age and symptom profile criteria. Engagement and retention procedures Identified candidates are contacted via a standardised protocol (telephone or in-person) to assess preliminary interest. Participants receive a Participant Information Sheet and Informed Consent form by email prior to the first visit to support informed decision-making. Monitoring Recruitment rates will be reviewed weekly. If enrolment falls below the projected target (approximately 15 participants per month), additional engagement with referring clinicians and expanded database screening will be undertaken Assignment of interventions: randomisation Sequence generation The randomisation schedule and block generation is performed independently by the Champalimaud Clinical Trials Unit (CTU), to ensure the sequence is not accessible to the study’s team. Type of randomisation Participants are randomly assigned to one of three parallel groups in a 1:1:1 allocation ratio using stratified permuted block randomisation with variable block sizes to ensure unpredictability and allocation balance throughout the recruitment period. Stratification is based on participants' current therapeutic status: List A (Non-ERP): participants who have not received Exposure and Response Prevention therapy in the six months prior to the study; List B (ERP-Ongoing): participants currently undergoing a stable psychotherapeutic process involving exposure work. Allocation concealment Allocation concealment is ensured by centralizing the randomisation process. The Master Randomisation Log is maintained exclusively by the CTU at CCC. The on-site research team, including the Principal Investigator, SLC, SSC and ERP Therapists, do not have access to the randomisation sequence or the block sizes. The allocation for each participant is only revealed to the SLC after the participant completed the baseline assessment. Implementation Enrolment of participants is conducted by the SLC. After baseline assessment (Visit 1), the SLC sends a standardised randomisation request to the CTU, specifying participant ID and stratum (List A or List B). The CTU identifies the next allocation and informs the SLC. Blinded ERP Therapists remain unaware of allocation throughout the study. Blinding Who is blinded This study uses an assessor-blind design. ERP Therapists and statisticians are blinded to group allocation. Treatment arms are coded as Group 1, Group 2, and Group 3 until all primary analyses are completed. How will blinding be achieved Assessments are conducted by the SLC and SSC, who are independent from the exposure sessions delivered by the ERP therapists. Participants are reminded not to disclose allocation. Session duration, frequency, structure, and therapist contact time are identical across all groups to ensure procedural consistency and minimise bias. Procedure for unblinding if needed Unblinding occurs only if essential for participant safety or clinical management (e.g. severe cybersickness). The principal investigator authorises unblinding through a formal request to the CTU. The date, reason, and personnel involved in any unblinding are documented in the trial master file. Unblinded participants remain in the study for follow-up assessments where clinically appropriate, with outcome ratings flagged accordingly. Data collection and management Assessment and collection of outcomes Most of the clinical assessments are self-report measures, with only two measures assessor-led (MINI and Y-BOCS-II). Prior to the start of recruitment, the SLC undertook training on how to administer these measures. Further details of assessment instruments are reported in Table 2 and Additional File 2. Retention and follow-up Reminders are sent via text and email the day before sessions. For consecutive sessions, next sessions are confirmed in person. Google Calendar invites with session details are emailed for all sessions. ERP Therapists and SLC or SSC check in with participants at the end of each session. Participants receive 25€ per visit (up to 175€ if they attend all seven visits), to compensate for travel and other participation costs. Participants may choose to complete only assessments or the final ERP session if they opt out of exposure. Data management Data entry and capture . Data collection uses both paper and electronic methods. Eligibility assessments (MINI 5.0, Y-BOCS-II) are recorded on paper, essential scores entered into the Participants Master Log, and originals archived securely. Self-report data are captured electronically via Qualtrics and a locked-down Microsoft Access database, with validation rules and an audit trail. Data coding and standardisation . Electronic self-report instruments require complete responses. SLC or SSC verify completion at the end of each session. Adverse events are recorded in free-text in the AE Log. Data quality and validation . BIOPAC physiological data undergo automated pre-processing and manual validation to remove artifacts. For daily quality control, synchronisation is achieved via post-hoc clock matching. That is, scene events are timestamped and sent as digital triggers/markers into BIOPAC (or logged and aligned). Data security and storage . The Master Randomisation Log is secured by the CTU. Data are transferred via hardline to the institutional server, backups are encrypted. 3D environments can be regenerated from JSON metadata if needed. Documentation . A comprehensive Data Dictionary (Codebook) and a README file detailing the directory structure of the AI assets, linking specific file paths to study arms and participant IDs. Confidentiality All study-related information is stored securely at the recruiting study site (CCC). To ensure confidentiality, all participant data is pseudonymised. The Master Randomisation Log is stored separately on a secure, password-protected institutional server. Physical documents are locked in filing cabinets; digital data are encrypted on dedicated laptops and servers. Data will be retained for 25 years in accordance with institutional and EU regulations. Statistical methods Primary outcome analyses Feasibility outcomes (recruitment, retention, data completeness, adherence) will be reported descriptively as proportions. Secondary outcome analyses Clinical Secondary outcomes will be analysed using a mixed effects linear model adjusted for baseline score and accounting for repeated measures with a random effect for participant. Adjusted between group mean differences will be presented alongside 95% confidence intervals (without p-values). Population for analysis All analyses follow the Intention-to-Treat (ITT) principle. Per-Protocol Sensitivity Analysis can include highly adherent participants. Handling missing data For primary feasibility outcomes, data will be reported descriptively for all randomised participants. Valid percentages will be calculated based on available data, and the extent and pattern of missing data will be clearly documented. For secondary outcomes involving repeated measurements, missing data will be handled using mixed effects models. These models utilize all available data from each participant under the missing at random assumption and will not be imputed. Qualitative Data Analysis The audio-recorded interviews will be transcribed, anonymised, and analysed using thematic analysis to identify themes related to acceptability, perceived utility, and barriers to the implementation of AI-generated VR exposure. Transparency and Reproducibility Analysis code, including data preprocessing scripts, model specifications, and visualization routines, will be documented and version-controlled during the study and will be made available upon reasonable request, subject to institutional approvals and any applicable data protection or intellectual property constraints. Protocol and statistical analysis plan This protocol contains the complete statistical analysis plan. No separate document will be submitted. Oversight and monitoring Coordinating centres and trial steering committee This study is led by King’s College London and Champalimaud Foundation with the latter as study sponsor and recruitment site. These centres oversee the day-to-day conduct of the trial. The Project Management Group (PMG) includes the chief investigator, principal investigator, SLC, trial manager, statisticians, and machine learning engineers. Members are based at King’s College London and at the Champalimaud Foundation. The PMG is responsible for overseeing trial progress, troubleshooting issues, and making operational decisions where required. The Trial Steering Committee (TSC) consists of independent members with expertise in psychiatry, psychology, VR, AI, and statistics. The TSC oversees the trial on behalf of the sponsor and funder, reviews relevant information from other sources, and advises the Chief Investigator, Principal Investigator and Sponsor on dissemination and presentation of the trial. Frequency and plans for auditing trial conduct Trial conduct is monitored by the TSC. Protocol amendments Should amendments be required, they will first be submitted to the REC for approval and then communicated to the Sponsor. Any consequent changes will be communicated to the site and, where needed, to participants. Dissemination policy Results will be disseminated via publications in peer-reviewed journals, conference presentations, and institutional websites (e.g. King’s College London and Champalimaud Foundation). Discussion This feasibility trial investigates whether GenAI-driven, personalised virtual reality (VR) exposure can be delivered safely and is acceptable for adults with obsessive–compulsive disorder (OCD). By evaluating recruitment, retention, adherence, data completeness, safety and participant experience, the study addresses key uncertainties that must be resolved before undertaking a future randomised controlled trial. A central innovation of this approach lies in standardising the process of exposure content generation while personalising the stimuli themselves . This may help to resolve a longstanding methodological tension in exposure-based interventions, particularly for OCD, where treatment protocols require consistency and replicability, yet therapeutic efficacy depends on close alignment between exposure stimuli and an individual’s idiosyncratic fears. Generative artificial intelligence enables the rapid creation of tailored exposure environments within a structured and reproducible framework, offering a potential solution to this challenge. Unlike earlier studies, which relied on fixed virtual libraries (26,27) or labour-intensive 360° video capture (28) , this approach allows therapists to co-create unique VR scenarios from a simple text prompt and optional image, tailored to the patient's idiosyncratic obsessions, and deployable on consumer-grade headsets. As a feasibility trial, it will yield critical information on recruitment pathways, participant burden, intervention adherence, safety monitoring procedures, and data collection methods. These data can inform refinement of the intervention and study procedures, including optimisation of exposure schedules, assessment timing, and selection of primary outcomes. Estimates of variability derived from secondary measures can support sample size estimation for a future randomised controlled trial. In this feasibility trial, the inclusion of both immersive (VR) and non-immersive (screen-based) delivery modes allows to compare the role of immersion, spatial presence, and embodied interaction in eliciting subjective distress and physiological responses. If shown to be feasible, acceptable and safe, this methodology may have important implications for the scalability and accessibility of exposure and response prevention (ERP). Traditional ERP often requires substantial therapist time and repeated in-person sessions, which can limit access and increase costs. If feasibility is demonstrated, this methodology has the potential to enhance the ecological validity, scalability, and efficiency of exposure-based treatments while preserving the core therapeutic principles of ERP. Trial status Recruitment for the trial started in January 2026. Abbreviations 3DGS: 3D Gaussian Splatting; AE: Adverse Event; CBT: Cognitive-Behavioral Therapy; CCC: Champalimaud Clinical Centre; CRF: Case Report Form; CTU: Clinical Trials Unit; ECG: Electrocardiogram; EDA: Electrodermal Activity; EDC: Electronic Data Capture; ERP: Exposure and Response Prevention; GDPR: General Data Protection Regulation; GenAI: Artificial Intelligence; ICD-10: International Classification of Diseases 10th Revision; JSON: JavaScript Object Notation; MINI: Mini International Neuropsychiatric Interview; OCD: Obsessive-Compulsive Disorder; OPP: Ordem dos Psicólogos Portugueses (Portuguese Psychologists Association); PPI: Patient and Public Involvement; RCT: Randomized Controlled Trial; REC: Research Ethics Committee; SAE: Serious Adverse Event; SPIRIT: Standard Protocol Items: Recommendations for Interventional Trials; SSQp: Simulator Sickness Questionnaire (Portuguese Version); STAI-Y1: State-Trait Anxiety Inventory (State version); SUDS: Subjective Units of Distress Scale; TSC: Trial Steering Committee; VAS: Visual Analogue Scale; VR: Virtual Reality VRET: Virtual Reality Exposure Therapy; Y-BOCS-II: Yale-Brown Obsessive-Compulsive Scale Second Edition. Declarations Acknowledgements The authors gratefully acknowledge Prof Thomas Craig, Prof Lucia Valmaggia, Dr Brennan Kahan, and Dr Carla Branco for their valuable contributions as members of the Trial Steering Committee. All individuals named in this Acknowledgements section have provided permission to be acknowledged. Authors’ contributions AA contributed to study design, participant recruitment, data collection, and writing, reviewing, and editing of the manuscript. BC is the senior statistician, supervised the development of the statistical analysis plan, and assisted with study design decisions. CF contributed to the trial management and setup, and to writing, reviewing and editing the manuscript. MPC is the Chief Investigator in the UK, contributed to securing study funding, study design, trial management, supervision, and writing, reviewing, and editing of the manuscript. JG contributed to study design. JK contributed to study design. ML contributed to the development of the statistical analysis plan and contributed to writing, reviewing, and editing of the manuscript. AL designed and deployed the end-to-end AI-driven pipeline for both VR and screen-based delivery. FM designed and deployed the end-to-end AI-driven pipeline for both VR and screen-based delivery. AJO-M is the Principal Investigator in Portugal, contributed to securing study funding, study design, trial management, supervision, and reviewing and editing of the manuscript. JS implemented the generative workflow for stimulus generation. All authors read and approved the final manuscript. Sources of funding and other support This research, MPC, CF, ML, AA and AJO-M are funded by the Department for Science, Innovation and Technology (DSIT). This study has been delivered with support from the National Institute for Health and Care Research (NIHR) Maudsley Biomedical Research Centre (BRC), who provides funding for MPC salary. AJO-M is supported by a Starting Grant (grant agreement no. 950357) and a Proof-of-Concept Grant (grant agreement no. 101158262) from the European Research Council, and by the PsyPal project (grant agreement no. 875358), all funded by the European Union’s Horizon 2020 Research and Innovation Programme. The funders did not have a role in the collection, management, analysis, interpretation of data, and in writing the manuscript. The views expressed are those of the author(s) and not necessarily those of the DSIT, the NIHR, the Department of Health and Social Care, the European Research Council or the European Union’s Horizon 2020 Research and Innovation Programme . Availability of data and materials The datasets generated and analysed during the current study will be available from the corresponding author on reasonable request. Ethics approval and consent to participate The study was approved by the Champalimaud Foundation’s Ethics Committee (07/10/2025). Informed consent is collected from all study participants. Consent for publication Not applicable as no data was included. Competing interests AJO-M is coordinating investigator of a multicentre trial for obsessive compulsive disorder, sponsored by Bioprojet Pharma (EUCT 2025-522026-13-00, protocol P24-05), was recipient of a grant from Schuhfried GmBH for norming and validation of cognitive tests, and national coordinator for Portugal of trials for treatment-resistant depression, sponsored by Compass Pathways, Ltd (EudraCT 2017-003288-36 and 2020-001348-25) and Janssen-Cilag, Ltd (EudraCT 2019-002992-33); has received payment, honoraria, or support for attending meetings and participating in advisory boards from MSD, Neurolite AG, Janssen Pharmaceuticals, Angelini Pharma, and the European Monitoring Centre for Drugs and Drug Addiction; has received consultancy fees from Bioprojet Pharma and NaturalX Health Ventures (all outside the submitted work); is President of the Scientific Council of the Portuguese Obsessive Compulsive Disorder Foundation. The remaining authors declare that they have no competing financial or other interests. No financial or other conflicts of interest are declared by members of the trial steering committee. References Australian Psychological Society. Evidence-based Psychological Interventions. 2018. Report. NICE. Obsessive-compulsive disorder and body dysmorphic disorder: treatment Clinical guideline [Internet]. 2005. Report. Available from: www.nice.org.uk/guidance/cg31 Foa EB, McLean CP. The Efficacy of Exposure Therapy for Anxiety-Related Disorders and Its Underlying Mechanisms: The Case of OCD and PTSD. Annu Rev Clin Psychol. 2016 Mar 28;12:1–28. doi:10.1146/annurev-clinpsy-021815-093533 PubMed PMID: 26565122. Blakey SM, Abramowitz JS. Interoceptive Exposure: An Overlooked Modality in the Cognitive-Behavioral Treatment of OCD The Nature and Treatment of OCD [Internet]. 2018. Report. Available from: www.elsevier.com/locate/cabp Bragdon LB, Eng GK, Belanger A, Collins KA, Stern ER. Interoception and Obsessive-Compulsive Disorder: A Review of Current Evidence and Future Directions. Frontiers in Psychiatry. Frontiers Media S.A.; 2021. doi:10.3389/fpsyt.2021.686482 Abramowitz JS. The Psychological Treatment of Obsessive-Compulsive Disorder. In Review Can J Psychiatry. 2006. Report. Pozza A, Dèttore D. Drop-out and efficacy of group versus individual cognitive behavioural therapy: What works best for Obsessive-Compulsive Disorder? A systematic review and meta-analysis of direct comparisons. Psychiatry Research. Elsevier Ireland Ltd; 2017. p. 24–36. doi:10.1016/j.psychres.2017.09.056 PubMed PMID: 28982038. Johnco C, McGuire JF, Roper T, Storch EA. A meta-analysis of dropout rates from exposure with response prevention and pharmacological treatment for youth with obsessive compulsive disorder. Depression and Anxiety. Blackwell Publishing Inc.; 2020. p. 407–17. doi:10.1002/da.22978 PubMed PMID: 31778595. Ong CW, Clyde JW, Bluett EJ, Levin ME, Twohig MP. Dropout rates in exposure with response prevention for obsessive-compulsive disorder: What do the data really say? Journal of Anxiety Disorders. Elsevier Ltd; 2016. p. 8–17. doi:10.1016/j.janxdis.2016.03.006 PubMed PMID: 27061971. Hezel DM, Simpson HB. Exposure and response prevention for obsessive‑compulsive disorder: A review and new directions. Indian J Psychiatry. 2019;61. Faustino D, Braga R, Faria M, Gonçalves M, Oliveira J. A Systematic Review on How to Combine Exposure and Response Prevention With Add-Ons for the Treatment of Obsessive–Compulsive Disorder. 2025. doi:10.1037/pst0000560.supp Wheaton MG, Chen SR. Homework Completion in Treating Obsessive–Compulsive Disorder with Exposure and Ritual Prevention: A Review of the Empirical Literature. Cognitive Therapy and Research. Springer; 2021. p. 236–49. doi:10.1007/s10608-020-10125-0 Hamatani S, Tsuchiyagaito A, Nihei M, Hayashi Y, Yoshida T, Takahashi J, et al. Predictors of response to exposure and response prevention-based cognitive behavioral therapy for obsessive-compulsive disorder. BMC Psychiatry. 2020 Sep 4;20(1). doi:10.1186/s12888-020-02841-4 PubMed PMID: 32887553. Bouchard S, Dumoulin S, Robillard G, Guitard T, Klinger E, Forget H, et al. Virtual reality compared with in vivo exposure in the treatment of social anxiety disorder: A three-arm randomised controlled trial. British Journal of Psychiatry. 2017 Apr 1;210(4):276–83. doi:10.1192/bjp.bp.116.184234 PubMed PMID: 27979818. Javaherirenani R, Mortazavi SS, Shalbafan M, Ashouri A, Farani AR. Virtual reality exposure and response prevention in the treatment of obsessive-compulsive disorder in patients with contamination subtype in comparison with in vivo exposure therapy: a randomized clinical controlled trial. BMC Psychiatry. 2022 Dec 1;22(1). doi:10.1186/s12888-022-04402-3 PubMed PMID: 36443695. Miegel F, Jelinek L, Lohse L, Moritz S, Blömer J, Juckoff K, et al. Exposure Therapy in Mixed Reality for Obsessive-Compulsive Disorder A Randomized Clinical Trial. JAMA Netw Open. 2025 May 1;8(5). doi:10.1001/jamanetworkopen.2025.11488 PubMed PMID: 40392553. Kim J, Pacheco JPG, Golden A, Aboujaoude E, van Roessel P, Gandhi A, et al. Artificial Intelligence in Obsessive-Compulsive Disorder: A Systematic Review. Current Treatment Options in Psychiatry. Springer Science and Business Media Deutschland GmbH; 2025. doi:10.1007/s40501-025-00359-8 Dehghan B, Saeidimehr S, Sayyah M, Rahim F. The Effect of Virtual Reality on Emotional Response and Symptoms Provocation in Patients With OCD: A Systematic Review and Meta-Analysis. Frontiers in Psychiatry. Frontiers Media S.A.; 2022. doi:10.3389/fpsyt.2021.733584 PubMed PMID: 35177996. van Loenen I, Scholten W, Muntingh A, Smit J, Batelaan N. The Effectiveness of Virtual Reality Exposure–Based Cognitive Behavioral Therapy for Severe Anxiety Disorders, Obsessive-Compulsive Disorder, and Posttraumatic Stress Disorder: Meta-analysis. Journal of Medical Internet Research. JMIR Publications Inc.; 2022. doi:10.2196/26736 PubMed PMID: 35142632. Freitas JRS, Velosa VHS, Abreu LTN, Jardim RL, Santos JAV, Peres B, et al. Virtual Reality Exposure Treatment in Phobias: a Systematic Review. Psychiatric Quarterly. 2021 Dec 1;92(4):1685–710. doi:10.1007/s11126-021-09935-6 PubMed PMID: 34173160. Colman M, Millar J, Patil B, Finnegan D, Russell A, Higson-Sweeney N, et al. A systematic review and narrative synthesis of the use and effectiveness of extended reality technology in the assessment, treatment and study of obsessive compulsive disorder. Journal of Obsessive-Compulsive and Related Disorders. Elsevier B.V.; 2024. doi:10.1016/j.jocrd.2024.100893 Chan AW, Boutron I, Hopewell S, Moher D, Schulz KF, Collins GS, et al. SPIRIT 2025 statement: updated guideline for protocols of randomised trials. BMJ. 2025 Apr 28;389:e081477. doi:10.1136/bmj-2024-081477 Gonçalves G, Melo M, Serodio C, Silva R, Bessa M. Adaptation and Validation of the Simulator Sickness Questionnaire to Portuguese (SSQp) Based on Immersive Virtual Reality Exposure. IEEE Access. 2024;12:92708–17. doi:10.1109/ACCESS.2024.3419589 Silva DR, Campos RC. Alguns dados normativos do Inventário de Estado-Traço de Ansiedade – Forma Y (STAI-Y), de Spielberger, para a população portuguesa. . Revista Portuguesa de Psicologia. 1998;33:71–89. Montgomery R. Sample size justification in feasibility studies: moving beyond published guidance. Pilot and Feasibility Studies. BioMed Central Ltd; 2025. doi:10.1186/s40814-025-01675-9 Cullen AJ, Dowling NL, Segrave R, Carter A, Yücel M. Exposure therapy in a virtual environment: Validation in obsessive compulsive disorder. J Anxiety Disord. 2021 May 1;80. doi:10.1016/j.janxdis.2021.102404 PubMed PMID: 33894550. Miegel F, Jelinek L, Lohse L, Moritz S, Blömer J, Juckoff K, et al. Exposure Therapy in Mixed Reality for Obsessive-Compulsive Disorder A Randomized Clinical Trial. JAMA Netw Open. 2025 May 1;8(5). doi:10.1001/jamanetworkopen.2025.11488 PubMed PMID: 40392553. Benzina N, Morgiève M, Euvrard M, Flores Alves Dos Santos J, Mallet L. Personalised 360 o video exposure therapy for the treatment of obsessive-compulsive 1 disorder: a single case study 2. French Journal of Psychiatry. 2020;1:31–8. doi:10.1016/j.fjpsy.2020.02.004 Table Table 2 is available in the Supplementary Files section. Additional Declarations The authors declare potential competing interests as follows: AJO-M is coordinating investigator of a multicentre trial for obsessive compulsive disorder, sponsored by Bioprojet Pharma (EUCT 2025-522026-13-00, protocol P24-05), was recipient of a grant from Schuhfried GmBH for norming and validation of cognitive tests, and national coordinator for Portugal of trials for treatment-resistant depression, sponsored by Compass Pathways, Ltd (EudraCT 2017-003288-36 and 2020-001348-25) and Janssen-Cilag, Ltd (EudraCT 2019-002992-33); has received payment, honoraria, or support for attending meetings and participating in advisory boards from MSD, Neurolite AG, Janssen Pharmaceuticals, Angelini Pharma, and the European Monitoring Centre for Drugs and Drug Addiction; has received consultancy fees from Bioprojet Pharma and NaturalX Health Ventures (all outside the submitted work); is President of the Scientific Council of the Portuguese Obsessive Compulsive Disorder Foundation. The remaining authors declare that they have no competing financial or other interests. No financial or other conflicts of interest are declared by members of the trial steering committee. Supplementary Files Table2.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9244535","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Method Article","associatedPublications":[],"authors":[{"id":613286516,"identity":"0aa59532-89ea-4cfa-8697-d0c74a728540","order_by":0,"name":"Mariana Pinto da Costa","email":"data:image/png;base64,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","orcid":"","institution":"King's College London","correspondingAuthor":true,"prefix":"","firstName":"Mariana","middleName":"Pinto da","lastName":"Costa","suffix":""},{"id":613286517,"identity":"10ab8df2-e31e-452f-ad78-2326bc713b1b","order_by":1,"name":"André Abreu","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"André","middleName":"","lastName":"Abreu","suffix":""},{"id":613286518,"identity":"f22308d7-7010-4f7a-8110-73f97256472e","order_by":2,"name":"Carolina Fialho","email":"","orcid":"","institution":"King’s College London","correspondingAuthor":false,"prefix":"","firstName":"Carolina","middleName":"","lastName":"Fialho","suffix":""},{"id":613286519,"identity":"8faf2e06-d266-4f19-84d7-59e2cbbe17ca","order_by":3,"name":"Maximin Lange","email":"","orcid":"","institution":"King’s College London","correspondingAuthor":false,"prefix":"","firstName":"Maximin","middleName":"","lastName":"Lange","suffix":""},{"id":613286520,"identity":"b44dbac2-531a-4958-80c6-c4740979297e","order_by":4,"name":"Ben Carter","email":"","orcid":"","institution":"King’s College London","correspondingAuthor":false,"prefix":"","firstName":"Ben","middleName":"","lastName":"Carter","suffix":""},{"id":613286521,"identity":"67ced82d-049e-4745-a510-6d129debc0dd","order_by":5,"name":"Jaime Grácio","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Jaime","middleName":"","lastName":"Grácio","suffix":""},{"id":613286522,"identity":"24a0d9a2-e19a-4525-8bf0-5dd1f569f8f4","order_by":6,"name":"Alexander Loktyushin","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Alexander","middleName":"","lastName":"Loktyushin","suffix":""},{"id":613286523,"identity":"bc7c1233-9248-4170-9dc3-a21f7732fbcf","order_by":7,"name":"Fatemeh Molaei","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Fatemeh","middleName":"","lastName":"Molaei","suffix":""},{"id":613286524,"identity":"92267ae6-dce1-46a5-8c33-bcf615c9b8b6","order_by":8,"name":"Johannes Stelzer","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Johannes","middleName":"","lastName":"Stelzer","suffix":""},{"id":613286525,"identity":"993c1876-acad-49e8-a2b2-4a5d690c25bb","order_by":9,"name":"John Krakauer","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"","lastName":"Krakauer","suffix":""},{"id":613286526,"identity":"e084e3a4-f5a0-4ada-a816-f01ac94d8f53","order_by":10,"name":"Albino J. Oliveira-Maia","email":"data:image/png;base64,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","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":true,"prefix":"","firstName":"Albino","middleName":"J.","lastName":"Oliveira-Maia","suffix":""}],"badges":[],"createdAt":"2026-03-27 12:06:03","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":true,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9244535/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9244535/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105874962,"identity":"779dbdf1-1866-49a1-9993-03d28caf4a29","added_by":"auto","created_at":"2026-04-01 05:33:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":785710,"visible":true,"origin":"","legend":"\u003cp\u003e(A) Example of OCD-related stimulus presented via a VR headset; (B) Example of neutral stimulus presented via a VR headset; (C) Example of OCD-related stimulus presented via a widescreen.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9244535/v1/c72cb6b2a5782812a26fe43d.png"},{"id":105874963,"identity":"f80b1a37-789e-4598-807e-b38af0fa3fb4","added_by":"auto","created_at":"2026-04-01 05:33:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":249112,"visible":true,"origin":"","legend":"\u003cp\u003eComponents in both experimental and control conditions\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9244535/v1/db9d3817787b93b96fa5da99.png"},{"id":106414695,"identity":"6c06f1ae-3a30-4b3e-b771-b6b874cbd788","added_by":"auto","created_at":"2026-04-08 10:22:47","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2709470,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9244535/v1/bb4ee689-8c85-47b4-ba7f-6a65c115cf3d.pdf"},{"id":105905081,"identity":"e4633f56-f92f-402c-9c04-7928f39b0293","added_by":"auto","created_at":"2026-04-01 10:11:25","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":80755,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-9244535/v1/bec60caf13de91a17549584f.docx"}],"financialInterests":"The authors declare potential competing interests as follows: AJO-M is coordinating investigator of a multicentre trial for obsessive compulsive disorder, sponsored by Bioprojet Pharma (EUCT 2025-522026-13-00, protocol P24-05), was recipient of a grant from Schuhfried GmBH for norming and validation of cognitive tests, and national coordinator for Portugal of trials for treatment-resistant depression, sponsored by Compass Pathways, Ltd (EudraCT 2017-003288-36 and 2020-001348-25) and Janssen-Cilag, Ltd (EudraCT 2019-002992-33); has received payment, honoraria, or support for attending meetings and participating in advisory boards from MSD, Neurolite AG, Janssen Pharmaceuticals, Angelini Pharma, and the European Monitoring Centre for Drugs and Drug Addiction; has received consultancy fees from Bioprojet Pharma and NaturalX Health Ventures (all outside the submitted work); is President of the Scientific Council of the Portuguese Obsessive Compulsive Disorder Foundation. The remaining authors declare that they have no competing financial or other interests. No financial or other conflicts of interest are declared by members of the trial steering committee.","formattedTitle":"\u003cp\u003eDeveloping immersive virtual exposures for obsessive-compulsive disorder – Protocol for a Randomised, Controlled, Proof-of-Concept Feasibility Trial\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCognitive-behavioural therapy (CBT), particularly exposure and response prevention (ERP), is the gold-standard treatment for obsessive-compulsive disorder (OCD) (1,2). ERP involves the gradual exposure to anxiety-provoking stimuli while refraining from compulsive behaviours intended to reduce distress. Exposure can be conducted in vivo, imaginally, or interoceptively (3\u0026ndash;5). Despite its efficacy, ERP is associated with several challenges that can limit its accessibility and uptake., While recent reviews report treatment dropout rates of approximately 10-16% (7\u0026ndash;9), earlier\u0026nbsp;reviews reported rates of up to 25-30%\u0026nbsp;\u003cw:sdt docpart=\"D23A085EA23C4FA985A9F58B51EBA56C\" sdttag=\"MENDELEY_CITATION_v3_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\" id=\"891387630\"\u003e(6)\u003c/w:sdt\u003e. Attrition has been attributed to the high emotional burden experienced by patients, particularly during early treatment phases, when the effort required to confront anxiety-provoking stimuli without performing their compulsive soothing rituals may seem harder to endure (10,11). ERP typically requires regular therapist contact, with frequent sessions and adherence to homework exposure tasks, commonly leading to scheduling difficulties and impose additional time and travel costs for patients (8,12,13). The clinical setting also imposes inherent constraints on treatment delivery, as therapists often lack the means to replicate the idiosyncratic, context-bound triggers for specific obsessions within the session confines, which risks diminishing ecological validity of the proposed exposure exercises and shifts the burden to the patient\u0026rsquo;s self-motivation and compliance (14).\u003c/p\u003e\n\u003cp\u003eGenerative Artificial Intelligence (GenAI) and virtual reality (VR) technologies may offer opportunities to address these barriers. These tools can facilitate personalised, immersive therapeutic interventions that may be delivered at scale and reduced cost\u0026nbsp;(15,16). GenAI models can simulate OCD-related stimuli in controlled environments, supporting the development of tailored exposure scenarios (17). Recent meta-analyses of virtual reality exposure therapy in anxiety and obsessive-compulsive spectrum disorders have reported mixed results (18\u0026ndash;20). A key limitation of existing VR interventions for OCD is their reliance on pre-built, standardised environments (e.g., dirty bathrooms, cluttered kitchens, or disorderly rooms) which are often difficult to adapt to the heterogeneous and idiosyncratic symptom presentations characteristic of the disorder. This is problematic given that VR exposures only elicit therapeutic responses when the stimuli closely match the individual\u0026rsquo;s fears and compulsions (21). This highlights the need to better integrate the idiosyncratic nature of effective ERP into a standardised delivery model, with individualised exposure content, ensuring that protocol rigidity does not come at the expense of ecological validity.\u003c/p\u003e\n\u003cp\u003eTo address this need, a novel approach standardising the process of creating exposure scenarios while personalising the scenarios to individual patients was developed. This methodology, using AI-driven stimulus creation with structured titration, allows protocol replication across sites, while maintaining the individualisation necessary for effective exposure therapy.\u003c/p\u003e\n\u003cp\u003eThis approach introduces several methodological and technical innovations over previous OCD VR and mixed reality research. It combines the rapid, on-demand generation of personalised, photorealistic 3D exposure environments using GenAI and Gaussian Splatting, thereby eliminating the need for pre-built assets, 3D modelling, or game engine development. However, it remains unclear whether AI-generated, individualised VR exposures are feasible and acceptable in clinical practice. Thus, this article reports the protocol of the DIVE-OCD (\u003cem\u003eDeveloping Immersive Virtual Exposures for OCD)\u003c/em\u003e randomised controlled feasibility trial.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eObjectives\u0026nbsp;\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe primary objective of this study is to evaluate the feasibility, acceptability, and safety of using generative artificial intelligence to produce individualised virtual reality exposure stimuli for patients with OCD. Secondary objectives are to collect data on the impact of the intervention in the physiological reactivity, cybersickness, general distress, and exposure distress. This study will explore emotional engagement with the exposure content when delivered via immersive virtual reality (VR) and via a screen-based format, as well as in comparison to neutral content.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003ePatient and public involvement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThree patient and public involvement (PPI) members, including two people with lived experience of OCD and one family member, contribute to the design, interpretation of findings and dissemination. During the design phase, PPI members review study materials to ensure language clarity and accessibility. During the analytic phase, they will review and help refine the wording of themes. During the reporting phase, they will assist in co-creating lay summaries, advise on dissemination strategies, and help communicate findings to local organisations and other individuals with lived experience. Two of these members also serve on the board of a Portuguese OCD association (\u003cem\u003ePOC Doentes e Fam\u0026iacute;lias),\u003c/em\u003e supporting the dissemination of evidence-based treatment options and resources for people with OCD.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTrial design\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis is a randomised controlled feasibility study with three parallel arms and single-blind (assessor-blinded) design, stratified by current therapy status (non-ERP versus ERP-ongoing). Participants are allocated in a 1:1:1 ratio to receive exposure via OCD-related VR environments, neutral VR environments, or OCD-related stimuli presented on a screen. It employs an exploratory framework\u0026nbsp;to evaluate the feasibility, acceptability and safety of GenAI-driven virtual reality technology as a tool to enhance adherence to exposure and response prevention therapy in OCD treatment. This protocol follows the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guideline (22) (see Additional file 1 for the SPIRIT checklist).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTrial setting\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants are recruited at the Champalimaud Clinical Centre (CCC), Lisbon, Portugal, a private hospital with a neuropsychiatry unit, and integrated research programs. All intervention sessions are delivered on-site. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParticipants\u0026rsquo; eligibility criteria \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInclusion criteria:\u003c/p\u003e\n\u003cul class=\"decimal_type\"\u003e\n \u003cli\u003eAdults aged 18 to 65 years\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePrimary diagnosis of OCD (ICD-10 F42.0, F42.1, F42.2) confirmed by Mini International Neuropsychiatric Interview (MINI 5.0.0)\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eY-BOCS-II score \u0026ge; 14 (moderate to extremely severe symptoms)\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAt least one OCD subtype amenable to visual exposure (e.g. contamination/washing, checking, or symmetry/ordering)\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAbility to provide informed consent\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eExclusion criteria:\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003ePredominant OCD symptoms not suitable for visual exposure\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eCurrent or past psychotic episode, schizophrenia, or autism spectrum disorder\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eCurrent substance use disorder\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSelf-reported history of severe cardiovascular disease where acute physiological arousal poses a safety risk, defined as the presence of an implanted cardiac device (pacemaker or ICD), a diagnosis of heart failure or aortic aneurysm, or a major cardiac event (myocardial infarction, stroke, or cardiac surgery) within the previous 12 months\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eVisual or auditory impairment, susceptibility to motion sickness, or epilepsy \u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eEligibility criteria for those delivering interventions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Lead Clinician (unblinded):\u003c/strong\u003e\u0026nbsp; clinical psychologist with more than two years of clinical experience and training in ERP for OCD; conducts baseline assessments, AI stimuli titration after training from machine learning experts, and supervises asynchronous exposure sessions (i.e. supervised exposure blocks/scenarios that are not therapist-led ERP). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Support Clinician (SSC) (unblinded):\u003c/strong\u003e psychologists or psychiatrists, who assist the SLC with recruitment, eligibility assessments, and monitoring asynchronous exposure sessions. SSC who are psychiatrists, provide immediate consultation for any urgent psychiatric care urgent needs arising during the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eERP Therapist (blinded):\u003c/strong\u003e\u0026nbsp; delivers the psychotherapeutic sessions (Visit 2A and Visit 7A) after receiving specific training on the manualised intervention. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed consent\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWritten informed consent is obtained by the research team following review of the participant information sheet and discussion of study procedures.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntervention and comparators descriptions\u0026nbsp;\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe trial includes three arms (Figure 1):\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003eOCD-VR arm: immersive personalised GenAI-generated VR environments corresponding to the participant\u0026rsquo;s specific OCD symptoms.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eNeutral-VR arm: immersive personalised GenAI-generated VR environments described by participants as non-threatening, e.g., natural landscapes.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eOCD-Screen arm: the same personalised GenAI-generated OCD-related stimuli as in the OCD-VR arm, delivered via a widescreen display. \u0026nbsp;\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eComparing AI-generated stimuli in VR versus widescreen isolates immersion effects, while OCD-VR versus Neutral-VR tests content-specific effects.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe intervention follows a sequential structure combining standard psychotherapeutic care with the experimental exposure component. Figure 2 displays the study components in both experimental and control conditions.\u003c/p\u003e\n\u003cp\u003eThe protocol is delivered over seven visits. Table 1 displays an overview of these visits. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 1. Overview of study visits and experimental procedures.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"650\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisit 1: Assessment and Titration\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eParticipants attend an initial session for diagnostic confirmation and stimulus generation. The Study Lead Clinician (SLC) or Study Support Clinician (SSC) conducts informed consent, confirms the OCD diagnosis and screens for possible diagnostic exclusion criteria using the Mini International Neuropsychiatric Interview (MINI 5.0). The SLC then assesses OCD severity and explores symptom description with the Y-BOCS-II, provides psychoeducation, and collaboratively constructs a personalised symptom hierarchy, ensuring all items fall within clinically appropriate intensity levels (SUDS \u0026le; 8).\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eDuring this session, participants also complete the AI-stimuli titration. They describe their triggers to generate sample images, which are iteratively refined with the SLC until clinically valid and within the target SUDS range. Individualised 2D images are generated using the FLUX.1 [dev] text-to-image model (Black Forest Labs, 2024) via the fal-ai/flux/dev endpoint using the fal-client Python library (v0.9.1 or higher) to ensure reproducibility. Prompt engineering prioritises detailed sensory descriptors over abstract concepts to ensure high ecological validity.\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eAll images undergo dual-layer safety review: automated model-based filters and manual inspection by the SLC to ensure appropriateness and therapeutic relevance. Images are only shown to participants after explicit approval by the SLC. Three categories of OCD-related stimuli and three categories of neutral images are finalised per participant. All generation parameters (prompt text, seed, model version, timestamps) are stored as JSON files, and all tasks are performed on a Lenovo LOQ Gaming 17i workstation to ensure consistent performance.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e3D Environment Generation\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eApproved 2D stimuli are used to generate immersive 3D environments using the Marble platform developed by World Labs (2024). Marble produces navigable volumetric scenes from image inputs using an GenAI-based 3D scene generation pipeline that represents environments with 3D Gaussian Splatting (3DGS). The resulting scenes preserve fine visual detail and allow real-time viewpoint changes. The generated scenes are exported and rendered using SparkJS for interactive viewing. VR exposures are delivered on Meta Quest 3 headsets with calibration for interpupillary distance and starting position. All scenes are pre-inspected by the Study Lead Clinician (SLC) to confirm visual quality and clinical suitability. Screen-based exposures are delivered via an 85-inch Neo QLED 4K display. The same 3DGS environments are used in both VR and screen conditions to ensure identical visual stimuli across modalities. The difference between modalities may contribute to outcomes. VR participants view the scene with natural 6DoF head tracking, while screen participants use a joystick to control camera orientation and navigation.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisits 2A and 2B:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eFirst ERP and First Asynchronous Session\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eApproximately one week after Visit 1, participants attend a 90-minute ERP session (Visit 2A) led by the blinded ERP Therapist. Exposure begins with lower-ranked items in the symptom hierarchy (progression interrupted if SUDS \u0026gt; 8). On the same day, participants complete the first of five daily 30-minute asynchronous sessions (Visit 2B) under supervision of the SLC or SSC. Physiological data (electrodermal activity and heart rate) are collected continuously, with subjective distress measured verbally at scene transitions.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisits 3\u0026ndash;6: Asynchronous Phase\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eParticipants complete four further 30-minute asynchronous exposure sessions on consecutive working days. Procedures, data collection, and supervision are identical to Visit 2B.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisit 7A:\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eSecond ERP Session\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eOne day after the final asynchronous phase visit, or at the next available opportunity within a 72-hour window, and approximately one week after Visit 2, participants attend a second 90-minute ERP session to progress further up the SUDS hierarchy under the guidance of the same blinded ERP Therapist.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisit 7B: Qualitative Debriefing\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 530px;\"\u003e\n \u003cp\u003eImmediately following Visit 7A, participants engage in a structured qualitative interview with the SLC to provide feedback on feasibility, acceptability, and experience of the intervention.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eStrategies to improve adherence to intervention/comparator:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn the first visit, the SLC or SSC explains the study procedures and answers any questions. All sessions and data collection occur in person, providing additional opportunities to encourage engagement. Participants receive daily text and email reminders for the upcoming appointments. If a session is missed, the SLC contacts the participant to reschedule the session and discusses any barriers to attendance.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCriteria for discontinuing or modifying allocated intervention/comparator\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThere is no participant relocation. If a participant is unable to continue with asynchronous exposure in VR or screen or experience any adverse events resulting in their wish to interrupt the asynchronous phase, their participation is discontinued from this intervention component. For an intention-to-treat analysis, these participants remain eligible for all subsequent outcome measures; specifically, they are invited to complete the second ERP visit and the qualitative debrief interview, unless data removal is explicitly requested.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConcomitant care permitted or prohibited during the trial:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants must maintain their current medication regimen as prescribed. Regarding psychotherapeutic interventions, participants are divided into two groups:\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eGroup A: Participants not currently receiving ERP must agree not to initiate any new ERP treatment during their participation in the study.\u003c/li\u003e\n \u003cli\u003eGroup B: Participants currently undergoing ERP, together with their treating therapists, must suspend all external ERP sessions and homework during the five consecutive-day asynchronous exposure phase (between the first and second study ERP sessions).\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eImmediate supportive care is available on-site for adverse events, namely if a participant experiences severe anxiety or motion sickness.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAncillary and post-trial care:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants are under outpatient psychiatric care in the CCC. Post-trial care is provided by their respective clinicians, who will be informed of any adverse events.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOutcomes:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe full assessment schedule is shown in Table 2.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrimary outcome\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003e\u003cstrong\u003eFeasibility:\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n \u003col start=\"1\" type=\"a\"\u003e\n \u003cli\u003e\u003cu\u003eRecruitment rate\u003c/u\u003e: number of participants enrolled within the recruitment window\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cu\u003eRetention rate\u003c/u\u003e: proportion of enrolled participants completing the final assessment\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cu\u003eData completeness\u003c/u\u003e: proportion of expected data points collected across all measures\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cu\u003eAdherence\u003c/u\u003e: proportion of planned exposure sessions completed in full (15 minutes)\u003c/li\u003e\n \u003c/ol\u003e\n\u003c/ol\u003e\n\u003col start=\"2\"\u003e\n \u003cli\u003e\u003cstrong\u003eSafety:\u003c/strong\u003e assessed through recording of any adverse events throughout the study period.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eAcceptability:\u003c/strong\u003e assessed through end of study qualitative interviews.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cstrong\u003eSecondary outcomes\u003c/strong\u003e\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003e\u003cstrong\u003ePhysiological reactivity:\u003c/strong\u003e measured by electrodermal activity (microsiemens) and heart rate (beats per minute) standardised within participants relative to within-session baseline.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCybersickness\u003c/strong\u003e: measured by the Simulator Sickness Questionnaire (SSQ) collected after each VR session (Portuguese version - SSQp; (23)).\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eGeneral distress:\u003c/strong\u003e measured by the visual analogue scale (0-10) and State-Trait Anxiety Inventory State version (STAI-Y1) collected after each exposure session (Portuguese version; (24)).\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eSubjective distress:\u003c/strong\u003e Change in the percent hierarchy position score (progression through the individualized SUDS hierarchy) from Session 2A (approximately 7 days post-randomisation) to Session 7A (approximately 14 days post-randomisation).\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cstrong\u003eAdverse Events\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAn \u003cu\u003eAdverse Event (AE)\u003c/u\u003e is any untoward medical or psychological occurrence, regardless of causal relationship with the study intervention. Given the nature of ERP, where anxiety induction is a therapeutic mechanism, a distinction is maintained between expected therapeutic distress and adverse events. High subjective distress (e.g., SUDS ratings \u0026gt;8/10) during exposure is considered therapeutic; however, distress that persists significantly beyond the session, panic attacks requiring medical support, or severe refusal to eat or drink due to symptom exacerbation are classified as AEs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA \u003cu\u003eSerious Adverse Event (SAE)\u003c/u\u003e is defined as any event that results in death, is life-threatening, requires inpatient hospitalisation, results in persistent or significant disability/incapacity, or requires intervention to prevent such outcomes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePotential harms anticipated in this study are consistent with those commonly associated with exposure therapy and VR research. Exposure to feared stimuli may temporarily increase anxiety or discomfort; to mitigate this risk, exposure hierarchies are constructed to exclude stimuli expected to elicit Subjective Units of Distress (SUDS) ratings above 8, and clinicians are present during all sessions to pause or terminate exposure as needed. Cybersickness or visual discomfort may occur in the VR condition and will be routinely monitored using the Simulator Sickness Questionnaire, with the option to interrupt or reschedule sessions if symptoms arise. The use of AI generation can introduce a theoretical risk of generating clinically inappropriate or disturbing content. To minimise this risk, all images will be screened through algorithmic safety guardrails and manually reviewed by the SLC before being shown to participants, ensuring that only clinically appropriate stimuli are used. Data and privacy risks are minimised through strict adherence to GDPR-compliant local storage: all immersive content will run on the VR or widescreen systems, and physiological recordings will be stored securely on institutional servers without cloud transfer. Collectively, these procedures are intended to identify, manage and document adverse events, supporting an assessment of the safety and feasibility of AI-driven immersive exposure within ERP for OCD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMonitoring and management\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSafety is monitored throughout the study by the unblinded SLC or SSC and the blinded ERP Therapist across the three arms. Specific intervention-related risks are managed via the following protocols:\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cu\u003eCybersickness\u003c/u\u003e: Visually induced motion sickness is monitored using the Simulator Sickness Questionnaire (SSQ) after each VR session. Acute symptoms (e.g., severe nausea) trigger immediate session cessation.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cu\u003eSuicidality\u003c/u\u003e: Monitored via clinical observation and the MINI interview. Any expression of suicidal ideation or intent triggers an immediate assessment by the on-site psychiatric team.\u003c/li\u003e\n \u003cli\u003e\u003cu\u003ePsychological distress\u003c/u\u003e: Sessions are terminated immediately upon participant request or if unmanageable distress (e.g., dissociation) is observed.\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eReporting and causality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll AEs will be recorded in the Case Report Form (CRF) and the AE log. Severity is graded on a three-tier scale (Mild, Moderate, Severe), and causality is assessed using a five-point scale (Not Related, Unlikely, Possible, Probable, Definitely). SAEs will be reported by the Principal Investigator to the on-site Research Ethics Committee (REC) within 15 days of the study team becoming aware of the event. In the case of urgent safety measures required to protect participants from immediate hazard, the REC will be notified within 3 days.\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParticipant timeline\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSee Table 2 for details on the participant schedule of enrolment, interventions, and assessments. Participants who are interested in taking part in the study attend the CCC for Visit 1, where they sign the informed consent form and complete the screening. If eligible, participants complete psychiatric diagnostic assessments, receive psychoeducation, collaboratively construct a personalised symptom hierarchy, and complete the stimulus preparation task with the SLC. Following this visit, participants are randomised. Participants return the following week for Visit 2A, the first ERP session with a clinical psychologist, followed by Visit 2B the same day, the first asynchronous session supervised by the SLC or SSC. Visits 3-6 occur the following 4 days and are similar asynchronous sessions. Participants then attend Visit 7A, the second ERP session, followed by session 7B, where participants engage in a qualitative interview with the SLC.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2. Participant timeline: schedule of enrolment, interventions, and assessments\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample size\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe target sample size for this feasibility trial is 45 participants, with 15 allocated to each of the three study arms. Following Montgomery (25), a sample of 15 per arm was chosen to prioritise the estimation of key parameters, specifically standard deviations and effect sizes, required to conduct a power analysis for a future RCT.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRecruitment\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParticipant identification and referral\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eProactive and multi-channel recruitment strategies are implemented, including:\u003c/p\u003e\n\u003cp\u003e(1) direct clinical referral from psychiatrists and psychologists within CCC\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(2) database systematic screening of the CCC electronic health records to identify patients with a primary diagnosis of OCD (ICD-10 F42.0, F42.1, F42.2) who meet age and symptom profile criteria.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEngagement and retention procedures\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIdentified candidates are contacted via a standardised protocol (telephone or in-person) to assess preliminary interest. \u0026nbsp;Participants receive a Participant Information Sheet and Informed Consent form by email prior to the first visit to support informed decision-making.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMonitoring\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRecruitment rates will be reviewed weekly. If enrolment falls below the projected target (approximately 15 participants per month), additional engagement with referring clinicians and expanded database screening will be undertaken\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssignment of interventions: randomisation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSequence generation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe randomisation schedule and block generation is performed independently by the Champalimaud Clinical Trials Unit (CTU), to ensure the sequence is not accessible to the study\u0026rsquo;s team.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eType of randomisation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants are randomly assigned to one of three parallel groups in a 1:1:1 allocation ratio using stratified permuted block randomisation with variable block sizes to ensure unpredictability and allocation balance throughout the recruitment period.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStratification is based on participants\u0026apos; current therapeutic status:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eList A (Non-ERP): participants who have not received Exposure and Response Prevention therapy in the six months prior to the study;\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eList B (ERP-Ongoing): participants currently undergoing a stable psychotherapeutic process involving exposure work.\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eAllocation concealment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAllocation concealment is ensured by centralizing the randomisation process. The Master Randomisation Log is maintained exclusively by the CTU at CCC. The on-site research team, including the Principal Investigator, SLC, SSC and ERP Therapists, do not have access to the randomisation sequence or the block sizes. The allocation for each participant is only revealed to the SLC after the participant completed the baseline assessment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImplementation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEnrolment of participants is conducted by the SLC. After baseline assessment (Visit 1), the SLC sends a standardised randomisation request to the CTU, specifying participant ID and stratum (List A or List B). The CTU identifies the next allocation and informs the SLC. Blinded ERP Therapists remain unaware of allocation throughout the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBlinding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWho is blinded\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study uses an assessor-blind design. ERP Therapists and statisticians are blinded to group allocation. Treatment arms are coded as Group 1, Group 2, and Group 3 until all primary analyses are completed.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHow will blinding be achieved\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAssessments are conducted by the SLC and SSC, who are independent from the exposure sessions delivered by the ERP therapists. Participants are reminded not to disclose allocation. Session duration, frequency, structure, and therapist contact time are identical across all groups to ensure procedural consistency and minimise bias.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProcedure for unblinding if needed\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUnblinding occurs only if essential for participant safety or clinical management (e.g. \u0026nbsp;severe cybersickness). The principal investigator authorises unblinding through a formal request to the CTU. The date, reason, and personnel involved in any unblinding are documented in the trial master file. Unblinded participants remain in the study for follow-up assessments where clinically appropriate, with outcome ratings flagged accordingly.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData collection and management\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssessment and collection of outcomes\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMost of the clinical assessments are self-report measures, with only two measures assessor-led (MINI and Y-BOCS-II). Prior to the start of recruitment, the SLC undertook training on how to administer these measures. Further details of assessment instruments are reported in Table 2 and Additional File 2.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRetention and follow-up\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eReminders are sent via text and email the day before sessions. For consecutive sessions, next sessions are confirmed in person. Google Calendar invites with session details are emailed for all sessions. ERP Therapists and SLC or SSC check in with participants at the end of each session. Participants receive 25\u0026euro; per visit (up to 175\u0026euro; if they attend all seven visits), to compensate for travel and other participation costs. Participants may choose to complete only assessments or the final ERP session if they opt out of exposure. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData management\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData entry and capture\u003c/strong\u003e.\u003cem\u003e\u0026nbsp;\u003c/em\u003eData collection uses both paper and electronic methods. Eligibility assessments (MINI 5.0, Y-BOCS-II) are recorded on paper, essential scores entered into the Participants Master Log, and originals archived securely. Self-report data are captured electronically via Qualtrics and a locked-down Microsoft Access database, with validation rules and an audit trail.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData coding and standardisation\u003c/strong\u003e. Electronic self-report instruments require complete responses. SLC or SSC verify completion at the end of each session. Adverse events are recorded in free-text in the AE Log.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData quality and validation\u003c/strong\u003e.\u003cem\u003e\u0026nbsp;\u003c/em\u003eBIOPAC physiological data undergo automated pre-processing and manual validation to remove artifacts. For daily quality control, synchronisation is achieved via post-hoc clock matching. That is, scene events are timestamped and sent as digital triggers/markers into BIOPAC (or logged and aligned).\u003cstrong\u003eData security and storage\u003c/strong\u003e. The Master Randomisation Log is secured by the CTU. Data are transferred via hardline to the institutional server, backups are encrypted. 3D environments can be regenerated from JSON metadata if needed.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDocumentation\u003c/strong\u003e.\u003cem\u003e\u0026nbsp;\u003c/em\u003eA comprehensive Data Dictionary (Codebook) and a README file detailing the directory structure of the AI assets, linking specific file paths to study arms and participant IDs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConfidentiality\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll study-related information is stored securely at the recruiting study site (CCC). To ensure confidentiality, all participant data is pseudonymised. The Master Randomisation Log is stored separately on a secure, password-protected institutional server. Physical documents are locked in filing cabinets; digital data are encrypted on dedicated laptops and servers. Data will be retained for 25 years in accordance with institutional and EU regulations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical methods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrimary outcome analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFeasibility outcomes (recruitment, retention, data completeness, adherence) will be reported descriptively as proportions.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eSecondary outcome analyses\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eClinical Secondary outcomes will be analysed using a mixed effects linear model adjusted for baseline score and accounting for repeated measures with a random effect for participant. Adjusted between group mean differences will be presented alongside 95% confidence intervals (without p-values).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePopulation for analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll analyses follow the Intention-to-Treat (ITT) principle. Per-Protocol Sensitivity Analysis can include highly adherent participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHandling missing data\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor primary feasibility outcomes, data will be reported descriptively for all randomised participants. Valid percentages will be calculated based on available data, and the extent and pattern of missing data will be clearly documented. For secondary outcomes involving repeated measurements, missing data will be handled using mixed effects models. These models utilize all available data from each participant under the missing at random assumption and will not be imputed.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQualitative Data Analysis\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe audio-recorded interviews will be transcribed, anonymised, and analysed using thematic analysis to identify themes related to acceptability, perceived utility, and barriers to the implementation of AI-generated VR exposure.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTransparency and Reproducibility\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnalysis code, including data preprocessing scripts, model specifications, and visualization routines, will be documented and version-controlled during the study and will be made available upon reasonable request, subject to institutional approvals and any applicable data protection or intellectual property constraints.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtocol and statistical analysis plan\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis protocol contains the complete statistical analysis plan. No separate document will be submitted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOversight and monitoring\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCoordinating centres and trial steering committee\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study is led by King\u0026rsquo;s College London and Champalimaud Foundation with the latter as study sponsor and recruitment site. \u0026nbsp;These centres oversee the day-to-day conduct of the trial. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe Project Management Group (PMG) includes the chief investigator, principal investigator, SLC, trial manager, statisticians, and machine learning engineers. Members are based at King\u0026rsquo;s College London and at the Champalimaud Foundation. The PMG is responsible for overseeing trial progress, troubleshooting issues, and making operational decisions where required.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe Trial Steering Committee (TSC) consists of independent members with expertise in psychiatry, psychology, VR, AI, and statistics. The TSC oversees the trial on behalf of the sponsor and funder, reviews relevant information from other sources, and advises the Chief Investigator, Principal Investigator and Sponsor on dissemination and presentation of the trial.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFrequency and plans for auditing trial conduct\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTrial conduct is monitored by the TSC.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtocol amendments\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eShould amendments be required, they will first be submitted to the REC for approval and then communicated to the Sponsor. Any consequent changes will be communicated to the site and, where needed, to participants.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDissemination policy\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eResults will be disseminated via publications in peer-reviewed journals, conference presentations, and institutional websites (e.g. King\u0026rsquo;s College London and Champalimaud Foundation).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis feasibility trial investigates whether GenAI-driven, personalised virtual reality (VR) exposure can be delivered safely and is acceptable for adults with obsessive\u0026ndash;compulsive disorder (OCD). By evaluating recruitment, retention, adherence, data completeness, safety and participant experience, the study addresses key uncertainties that must be resolved before undertaking a future randomised controlled trial. A central innovation of this approach lies in standardising the \u003cem\u003eprocess\u003c/em\u003e of exposure content generation while personalising the \u003cem\u003estimuli themselves\u003c/em\u003e. This may help to resolve a longstanding methodological tension in exposure-based interventions, particularly for OCD, where treatment protocols require consistency and replicability, yet therapeutic efficacy depends on close alignment between exposure stimuli and an individual\u0026rsquo;s idiosyncratic fears. Generative artificial intelligence enables the rapid creation of tailored exposure environments within a structured and reproducible framework, offering a potential solution to this challenge. Unlike earlier studies, which relied on fixed virtual libraries (26,27) or labour-intensive 360\u0026deg; video capture \u003cw:sdt docpart=\"0A9A5C6E288D4F74993E5329D1E6F3E9\" sdttag=\"MENDELEY_CITATION_v3_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\" id=\"937105356\"\u003e(28)\u003c/w:sdt\u003e, this approach allows therapists to co-create unique VR scenarios from a simple text prompt and optional image, tailored to the patient\u0026apos;s idiosyncratic obsessions, and deployable on consumer-grade headsets.\u003c/p\u003e\n\u003cp\u003eAs a feasibility trial, it will yield critical information on recruitment pathways, participant burden, intervention adherence, safety monitoring procedures, and data collection methods. These data can inform refinement of the intervention and study procedures, including optimisation of exposure schedules, assessment timing, and selection of primary outcomes. Estimates of variability derived from secondary measures can support sample size estimation for a future randomised controlled trial.\u003c/p\u003e\n\u003cp\u003eIn this feasibility trial, the inclusion of both immersive (VR) and non-immersive (screen-based) delivery modes allows to compare the role of immersion, spatial presence, and embodied interaction in eliciting subjective distress and physiological responses.\u003c/p\u003e\n\u003cp\u003eIf shown to be feasible, acceptable and safe, this methodology may have important implications for the scalability and accessibility of exposure and response prevention (ERP). Traditional ERP often requires substantial therapist time and repeated in-person sessions, which can limit access and increase costs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIf feasibility is demonstrated, this methodology has the potential to enhance the ecological validity, scalability, and efficiency of exposure-based treatments while preserving the core therapeutic principles of ERP.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTrial status\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRecruitment for the trial started in January 2026.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e\u003cstrong\u003e3DGS:\u003c/strong\u003e 3D Gaussian Splatting; \u003cstrong\u003eAE:\u003c/strong\u003e Adverse Event; \u0026nbsp;\u003cstrong\u003eCBT:\u0026nbsp;\u003c/strong\u003eCognitive-Behavioral Therapy; \u003cstrong\u003eCCC:\u003c/strong\u003e Champalimaud Clinical Centre; \u003cstrong\u003eCRF:\u003c/strong\u003e Case Report Form; \u003cstrong\u003eCTU:\u003c/strong\u003e Clinical Trials Unit; \u003cstrong\u003eECG:\u003c/strong\u003e Electrocardiogram; \u003cstrong\u003eEDA:\u003c/strong\u003e Electrodermal Activity; \u003cstrong\u003eEDC:\u003c/strong\u003e Electronic Data Capture; \u003cstrong\u003eERP:\u0026nbsp;\u003c/strong\u003eExposure and Response Prevention; \u003cstrong\u003eGDPR:\u003c/strong\u003e General Data Protection Regulation; \u003cstrong\u003eGenAI:\u003c/strong\u003e Artificial Intelligence;\u003cstrong\u003e\u0026nbsp;ICD-10:\u003c/strong\u003e International Classification of Diseases 10th Revision; \u003cstrong\u003eJSON:\u003c/strong\u003e JavaScript Object Notation; \u003cstrong\u003eMINI:\u003c/strong\u003e Mini International Neuropsychiatric Interview; \u003cstrong\u003eOCD:\u003c/strong\u003e Obsessive-Compulsive Disorder; \u003cstrong\u003eOPP:\u003c/strong\u003e Ordem dos Psic\u0026oacute;logos Portugueses (Portuguese Psychologists Association); \u003cstrong\u003ePPI:\u003c/strong\u003e Patient and Public Involvement; \u003cstrong\u003eRCT:\u003c/strong\u003e Randomized Controlled Trial; \u003cstrong\u003eREC:\u003c/strong\u003e Research Ethics Committee; \u003cstrong\u003eSAE:\u003c/strong\u003e Serious Adverse Event; \u003cstrong\u003eSPIRIT:\u003c/strong\u003e Standard Protocol Items: Recommendations for Interventional Trials; \u003cstrong\u003eSSQp:\u003c/strong\u003e Simulator Sickness Questionnaire (Portuguese Version); \u003cstrong\u003eSTAI-Y1:\u003c/strong\u003e State-Trait Anxiety Inventory (State version); \u003cstrong\u003eSUDS:\u003c/strong\u003e Subjective Units of Distress Scale; \u003cstrong\u003eTSC:\u003c/strong\u003e Trial Steering Committee; \u003cstrong\u003eVAS:\u003c/strong\u003e Visual Analogue Scale; \u003cstrong\u003eVR:\u003c/strong\u003e Virtual Reality \u003cstrong\u003eVRET:\u003c/strong\u003e Virtual Reality Exposure Therapy; \u003cstrong\u003eY-BOCS-II:\u003c/strong\u003e Yale-Brown Obsessive-Compulsive Scale Second Edition.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors gratefully acknowledge Prof Thomas Craig, Prof Lucia Valmaggia, Dr Brennan Kahan, and Dr Carla Branco for their valuable contributions as members of the Trial Steering Committee. All individuals named in this Acknowledgements section have provided permission to be acknowledged.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAA contributed to study design, participant recruitment, data collection, and writing, reviewing, and editing of the manuscript. BC is the senior statistician, supervised the development of the statistical analysis plan, and assisted with study design decisions. CF contributed to the trial management and setup, and to writing, reviewing and editing the manuscript. MPC is the Chief Investigator in the UK, contributed to securing study funding, study design, trial management, supervision, and writing, reviewing, and editing of the manuscript. JG contributed to study design. JK contributed to study design. ML contributed to the development of the statistical analysis plan and contributed to writing, reviewing, and editing of the manuscript. AL designed and deployed the end-to-end AI-driven pipeline for both VR and screen-based delivery. FM designed and deployed the end-to-end AI-driven pipeline for both VR and screen-based delivery. AJO-M is the Principal Investigator in Portugal, contributed to securing study funding, study design, trial management, supervision, and reviewing and editing of the manuscript. JS implemented the generative workflow for stimulus generation. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSources of funding and other support \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research, MPC, CF, ML, AA and AJO-M are funded by the Department for Science, Innovation and Technology (DSIT). This study has been delivered with support from the National Institute for Health and Care Research (NIHR) Maudsley Biomedical Research Centre (BRC), who provides funding for MPC salary. AJO-M is supported by a Starting Grant (grant agreement no. 950357) and a Proof-of-Concept Grant (grant agreement no. 101158262) from the European Research Council, and by the PsyPal project (grant agreement no. 875358), all funded by the European Union\u0026rsquo;s Horizon 2020 Research and Innovation Programme. The funders did not have a role in the collection, management, analysis, interpretation of data, and in writing the manuscript. The views expressed are those of the author(s) and not necessarily those of the DSIT, the NIHR, the Department of Health and Social Care, the European Research Council or the European Union\u0026rsquo;s Horizon 2020 Research and Innovation Programme\u003cem\u003e.\u003c/em\u003e\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analysed during the current study will be available from the corresponding author on reasonable request.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was approved by the Champalimaud Foundation\u0026rsquo;s Ethics Committee (07/10/2025). Informed consent is collected from all study participants. \u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable as no data was included.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eCompeting interests \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAJO-M is coordinating investigator of a multicentre trial for obsessive compulsive disorder, sponsored by Bioprojet Pharma (EUCT 2025-522026-13-00, protocol P24-05), was recipient of a grant from Schuhfried GmBH for norming and validation of cognitive tests, and national coordinator for Portugal of trials for treatment-resistant depression, sponsored by Compass Pathways, Ltd (EudraCT 2017-003288-36 and 2020-001348-25) and Janssen-Cilag, Ltd (EudraCT 2019-002992-33); has received payment, honoraria, or support for attending meetings and participating in advisory boards from MSD, Neurolite AG, Janssen Pharmaceuticals, Angelini Pharma, and the European Monitoring Centre for Drugs and Drug Addiction; has received consultancy fees from Bioprojet Pharma and NaturalX Health Ventures (all outside the submitted work); is President of the Scientific Council of the Portuguese Obsessive Compulsive Disorder Foundation. The remaining authors declare that they have no competing financial or other interests. No financial or other conflicts of interest are declared by members of the trial steering committee.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAustralian Psychological Society. Evidence-based Psychological Interventions. 2018. Report.\u003c/li\u003e\n\u003cli\u003eNICE. Obsessive-compulsive disorder and body dysmorphic disorder: treatment Clinical guideline [Internet]. 2005. Report. Available from: www.nice.org.uk/guidance/cg31\u003c/li\u003e\n\u003cli\u003eFoa EB, McLean CP. The Efficacy of Exposure Therapy for Anxiety-Related Disorders and Its Underlying Mechanisms: The Case of OCD and PTSD. Annu Rev Clin Psychol. 2016 Mar 28;12:1\u0026ndash;28. doi:10.1146/annurev-clinpsy-021815-093533 PubMed PMID: 26565122.\u003c/li\u003e\n\u003cli\u003eBlakey SM, Abramowitz JS. Interoceptive Exposure: An Overlooked Modality in the Cognitive-Behavioral Treatment of OCD The Nature and Treatment of OCD [Internet]. 2018. Report. Available from: www.elsevier.com/locate/cabp\u003c/li\u003e\n\u003cli\u003eBragdon LB, Eng GK, Belanger A, Collins KA, Stern ER. Interoception and Obsessive-Compulsive Disorder: A Review of Current Evidence and Future Directions. Frontiers in Psychiatry. Frontiers Media S.A.; 2021. doi:10.3389/fpsyt.2021.686482\u003c/li\u003e\n\u003cli\u003eAbramowitz JS. The Psychological Treatment of Obsessive-Compulsive Disorder. In Review Can J Psychiatry. 2006. Report.\u003c/li\u003e\n\u003cli\u003ePozza A, D\u0026egrave;ttore D. Drop-out and efficacy of group versus individual cognitive behavioural therapy: What works best for Obsessive-Compulsive Disorder? A systematic review and meta-analysis of direct comparisons. Psychiatry Research. Elsevier Ireland Ltd; 2017. p. 24\u0026ndash;36. doi:10.1016/j.psychres.2017.09.056 PubMed PMID: 28982038.\u003c/li\u003e\n\u003cli\u003eJohnco C, McGuire JF, Roper T, Storch EA. A meta-analysis of dropout rates from exposure with response prevention and pharmacological treatment for youth with obsessive compulsive disorder. Depression and Anxiety. Blackwell Publishing Inc.; 2020. p. 407\u0026ndash;17. doi:10.1002/da.22978 PubMed PMID: 31778595.\u003c/li\u003e\n\u003cli\u003eOng CW, Clyde JW, Bluett EJ, Levin ME, Twohig MP. Dropout rates in exposure with response prevention for obsessive-compulsive disorder: What do the data really say? Journal of Anxiety Disorders. Elsevier Ltd; 2016. p. 8\u0026ndash;17. doi:10.1016/j.janxdis.2016.03.006 PubMed PMID: 27061971.\u003c/li\u003e\n\u003cli\u003eHezel DM, Simpson HB. Exposure and response prevention for obsessive‑compulsive disorder: A review and new directions. Indian J Psychiatry. 2019;61.\u003c/li\u003e\n\u003cli\u003eFaustino D, Braga R, Faria M, Gon\u0026ccedil;alves M, Oliveira J. A Systematic Review on How to Combine Exposure and Response Prevention With Add-Ons for the Treatment of Obsessive\u0026ndash;Compulsive Disorder. 2025. doi:10.1037/pst0000560.supp\u003c/li\u003e\n\u003cli\u003eWheaton MG, Chen SR. Homework Completion in Treating Obsessive\u0026ndash;Compulsive Disorder with Exposure and Ritual Prevention: A Review of the Empirical Literature. Cognitive Therapy and Research. Springer; 2021. p. 236\u0026ndash;49. doi:10.1007/s10608-020-10125-0\u003c/li\u003e\n\u003cli\u003eHamatani S, Tsuchiyagaito A, Nihei M, Hayashi Y, Yoshida T, Takahashi J, et al. Predictors of response to exposure and response prevention-based cognitive behavioral therapy for obsessive-compulsive disorder. BMC Psychiatry. 2020 Sep 4;20(1). doi:10.1186/s12888-020-02841-4 PubMed PMID: 32887553.\u003c/li\u003e\n\u003cli\u003eBouchard S, Dumoulin S, Robillard G, Guitard T, Klinger E, Forget H, et al. Virtual reality compared with in vivo exposure in the treatment of social anxiety disorder: A three-arm randomised controlled trial. British Journal of Psychiatry. 2017 Apr 1;210(4):276\u0026ndash;83. doi:10.1192/bjp.bp.116.184234 PubMed PMID: 27979818.\u003c/li\u003e\n\u003cli\u003eJavaherirenani R, Mortazavi SS, Shalbafan M, Ashouri A, Farani AR. Virtual reality exposure and response prevention in the treatment of obsessive-compulsive disorder in patients with contamination subtype in comparison with in vivo exposure therapy: a randomized clinical controlled trial. BMC Psychiatry. 2022 Dec 1;22(1). doi:10.1186/s12888-022-04402-3 PubMed PMID: 36443695.\u003c/li\u003e\n\u003cli\u003eMiegel F, Jelinek L, Lohse L, Moritz S, Bl\u0026ouml;mer J, Juckoff K, et al. Exposure Therapy in Mixed Reality for Obsessive-Compulsive Disorder A Randomized Clinical Trial. JAMA Netw Open. 2025 May 1;8(5). doi:10.1001/jamanetworkopen.2025.11488 PubMed PMID: 40392553.\u003c/li\u003e\n\u003cli\u003eKim J, Pacheco JPG, Golden A, Aboujaoude E, van Roessel P, Gandhi A, et al. Artificial Intelligence in Obsessive-Compulsive Disorder: A Systematic Review. Current Treatment Options in Psychiatry. Springer Science and Business Media Deutschland GmbH; 2025. doi:10.1007/s40501-025-00359-8\u003c/li\u003e\n\u003cli\u003eDehghan B, Saeidimehr S, Sayyah M, Rahim F. The Effect of Virtual Reality on Emotional Response and Symptoms Provocation in Patients With OCD: A Systematic Review and Meta-Analysis. Frontiers in Psychiatry. Frontiers Media S.A.; 2022. doi:10.3389/fpsyt.2021.733584 PubMed PMID: 35177996.\u003c/li\u003e\n\u003cli\u003evan Loenen I, Scholten W, Muntingh A, Smit J, Batelaan N. The Effectiveness of Virtual Reality Exposure\u0026ndash;Based Cognitive Behavioral Therapy for Severe Anxiety Disorders, Obsessive-Compulsive Disorder, and Posttraumatic Stress Disorder: Meta-analysis. Journal of Medical Internet Research. JMIR Publications Inc.; 2022. doi:10.2196/26736 PubMed PMID: 35142632.\u003c/li\u003e\n\u003cli\u003eFreitas JRS, Velosa VHS, Abreu LTN, Jardim RL, Santos JAV, Peres B, et al. Virtual Reality Exposure Treatment in Phobias: a Systematic Review. Psychiatric Quarterly. 2021 Dec 1;92(4):1685\u0026ndash;710. doi:10.1007/s11126-021-09935-6 PubMed PMID: 34173160.\u003c/li\u003e\n\u003cli\u003eColman M, Millar J, Patil B, Finnegan D, Russell A, Higson-Sweeney N, et al. A systematic review and narrative synthesis of the use and effectiveness of extended reality technology in the assessment, treatment and study of obsessive compulsive disorder. Journal of Obsessive-Compulsive and Related Disorders. Elsevier B.V.; 2024. doi:10.1016/j.jocrd.2024.100893\u003c/li\u003e\n\u003cli\u003eChan AW, Boutron I, Hopewell S, Moher D, Schulz KF, Collins GS, et al. SPIRIT 2025 statement: updated guideline for protocols of randomised trials. BMJ. 2025 Apr 28;389:e081477. doi:10.1136/bmj-2024-081477\u003c/li\u003e\n\u003cli\u003eGon\u0026ccedil;alves G, Melo M, Serodio C, Silva R, Bessa M. Adaptation and Validation of the Simulator Sickness Questionnaire to Portuguese (SSQp) Based on Immersive Virtual Reality Exposure. IEEE Access. 2024;12:92708\u0026ndash;17. doi:10.1109/ACCESS.2024.3419589\u003c/li\u003e\n\u003cli\u003eSilva DR, Campos RC. Alguns dados normativos do Invent\u0026aacute;rio de Estado-Tra\u0026ccedil;o de Ansiedade \u0026ndash; Forma Y (STAI-Y), de Spielberger, para a popula\u0026ccedil;\u0026atilde;o portuguesa. . Revista Portuguesa de Psicologia. 1998;33:71\u0026ndash;89.\u003c/li\u003e\n\u003cli\u003eMontgomery R. Sample size justification in feasibility studies: moving beyond published guidance. Pilot and Feasibility Studies. BioMed Central Ltd; 2025. doi:10.1186/s40814-025-01675-9\u003c/li\u003e\n\u003cli\u003eCullen AJ, Dowling NL, Segrave R, Carter A, Y\u0026uuml;cel M. Exposure therapy in a virtual environment: Validation in obsessive compulsive disorder. J Anxiety Disord. 2021 May 1;80. doi:10.1016/j.janxdis.2021.102404 PubMed PMID: 33894550.\u003c/li\u003e\n\u003cli\u003eMiegel F, Jelinek L, Lohse L, Moritz S, Bl\u0026ouml;mer J, Juckoff K, et al. Exposure Therapy in Mixed Reality for Obsessive-Compulsive Disorder A Randomized Clinical Trial. JAMA Netw Open. 2025 May 1;8(5). doi:10.1001/jamanetworkopen.2025.11488 PubMed PMID: 40392553.\u003c/li\u003e\n\u003cli\u003eBenzina N, Morgi\u0026egrave;ve M, Euvrard M, Flores Alves Dos Santos J, Mallet L. Personalised 360 o video exposure therapy for the treatment of obsessive-compulsive 1 disorder: a single case study 2. French Journal of Psychiatry. 2020;1:31\u0026ndash;8. doi:10.1016/j.fjpsy.2020.02.004 \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 2 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Champalimaud Foundation","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Obsessive compulsive disorder, Virtual reality, Exposure and response prevention, Psychotherapy, Generative artificial intelligence, Feasibility study, Personalised treatment, 3D Gaussian Splatting, Immersive technology","lastPublishedDoi":"10.21203/rs.3.rs-9244535/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9244535/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eExposure and response prevention (ERP) is the gold-standard psychological treatment for obsessive compulsive disorder (OCD). However, its delivery typically requires frequent therapist availability and repeated patient travel to treatment settings, which limits accessibility. In addition, the idiosyncratic nature of obsessive-compulsive symptoms presents challenges for conducting effective exposures within the time and material constraints of traditional clinical environments. Virtual reality (VR)-based interventions may help address these limitations. This study aims to assess the feasibility of a novel approach that uses generative artificial intelligence (GenAI) to create personalised, immersive 3D exposure environments tailored to individual patient fears.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis is a randomised controlled feasibility study with three parallel arms and an assessor-blinded design. Forty-five adults with a primary diagnosis of OCD and moderate to extremely severe symptoms will be randomly allocated in a 1:1:1 ratio to OCD-related exposure administered via VR environments, neutral VR environments, or OCD-related stimuli presented on a widescreen display. Participants will complete a baseline assessment and an GenAI-based stimulus titration session, followed by two therapist-led ERP sessions that frame five consecutive days of asynchronous exposure (exposure blocks/scenarios that are not therapist-led ERP). The intervention uses text-to-image synthesis, image conversion into 3D Gaussian Splatting environments, and delivery via VR headsets or widescreen display. Primary feasibility outcomes include recruitment and retention rates, data completeness, and adherence to the asynchronous exposure protocol. Secondary outcomes include progression through personalised exposure hierarchies, physiological reactivity (electrodermal activity and heart rate), cybersickness, and subjective distress measures.\u003c/p\u003e\u003ch2\u003eDiscussion\u003c/h2\u003e \u003cp\u003eThis study will determine whether AI-driven VR exposure is feasible, safe, and acceptable for adults with OCD. The proposed approach standardises the process of content generation while personalising the actual stimuli, towards leveraging technology to ensure the personalised needs of these patients are more effectively met. Results can inform the refinement of the intervention and the study procedures, including sample size estimation for a future randomised controlled trial. If successful, this methodology could have the potential to improve scalability, reduce costs, and enhance the ecological validity of exposure therapy while maintaining clinical efficacy.\u003c/p\u003e\u003ch2\u003eTrial registration\u003c/h2\u003e \u003cp\u003eISRCTN13869986. Registered 29 December 2025. Prospectively registered.\u003c/p\u003e","manuscriptTitle":"Developing immersive virtual exposures for obsessive-compulsive disorder – Protocol for a Randomised, Controlled, Proof-of-Concept Feasibility Trial","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-01 05:33:05","doi":"10.21203/rs.3.rs-9244535/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"35f4af86-4401-4338-8db2-88f8d3beeae4","owner":[],"postedDate":"April 1st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T05:33:05+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-01 05:33:05","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9244535","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9244535","identity":"rs-9244535","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}