Mirror exposure therapy (MET) in virtual reality (VR) for body image disturbances (BIDS)

Abstract

Body Image Disturbances (BIDs) are increasingly prevalent, particularly among younger individuals influenced by social media and digital self‐comparison. Mirror Exposure Therapy (MET), a clinically supported intervention, promotes non‐judgemental, sustained observation of one’s body to reduce negative self‐perception. However, delivering MET in traditional clinical settings can be emotionally intense and logistically challenging. This project explores the feasibility of integrating MET with Virtual Reality (VR) and Artificial Intelligence (AI) to enhance personalisation, safety, and engagement while maintaining its therapeutic foundation. The aim was to develop a prototype that adapts MET to an immersive VR environment, supported by AI interaction and therapist supervision. A complete VR MET platform was built using Unity and deployed on the VIVE Focus Vision headset, selected for its built‐in eye tracking and standalone functionality. The system includes a 3D‐scanned avatar, AI‐guided prompts powered by a locally hosted Mistral 7B model via Ollama, real‐time transcription using Whisper, audio capture, and gaze and blink logging. Sessions take place in a virtual forest environment with camera switching and voice‐based interaction, overseen by a therapist who can manually pause the AI when needed. Qualitative evaluation involved two MET professionals, two former MET recipients, and one expert in VR‐based therapy. Participants reported high levels of user acceptance, perceived therapeutic realism, and valued features such as session transcripts and gaze tracking. They also highlighted emotional safety mechanisms, including phased avatar realism and delayed microphone activation. The findings support the feasibility and clinical potential of AI and VR supported MET when therapist oversight is retained. Limitations include technical complexity, small sample size, and inconsistencies in avatar quality. Future development should consider biometric inputs, live therapist dashboards, and longitudinal application. This work contributes to the growing field of immersive, ethically grounded digital mental health interventions.