# Online VR Diagnostic Platform (VCare) A therapist and a patient in the same virtual room. A UCSB capstone built with **WellHealth** by team **SegFault**: an exposure-therapy environment in Unity that adapts to the patient during the session, paired with a live console for the clinician. Second place, UCSB Capstone 2022. ## Product flow 1. A patient and a therapist join the same session — the patient in a VR headset, the therapist on the web client. 2. The patient lands in an environment built to contain the source of their anxiety: a social setting with people, ambient noise, and conversation in it. 3. Speech from the session is transcribed and run through tone and language analysis. 4. A feedback loop turns that analysis into environment parameters — crowd size, ambient volume, conversation intensity — so difficulty follows the patient rather than a fixed script. 5. The therapist sees what the patient sees, plus the transcript and session signals, and can override any parameter or take manual control. 6. After the session, the captured data is compiled into over-time graphs and a simplified score for both patient and therapist. ## System boundaries - **VR client (Unity XR)** — owns the environment, the avatars, the menu (room key, scene switch), and the local half of the level adjustment. Renders whatever game state the session says is current. - **Therapist console (web)** — owns the clinician's view: the video feed from the session, the live transcript, session signals, and the manual controls. - **Session services (Python/Django)** — own identity, session membership, the transcript pipeline and the server-side adjustment algorithm. The single authority on session state; both clients render it rather than deciding it. - **Cloud & data (AWS, MongoDB)** — deployment, the speech/analysis services, and the session and patient records the post-session analytics are computed from. ## Implementation choices - **Parameters, not scenes.** Anything therapeutic about the environment is expressed as an adjustable parameter with a level, so the same scene serves a first session and a tenth one, and so a therapist's manual override and the automatic loop act on the same controls. - **Server holds session state.** The adjustment algorithm runs server-side with a client-side counterpart, so the therapist's console and the patient's headset cannot disagree about the current level. - **Real-time first, analytics second.** The transcript pipeline exists for the live loop; the post-session graphs and score are computed from the records it leaves behind rather than from a separate capture path. ## Technology reference - **Experience:** Unity XR - **Services:** Python, Django - **Cloud & data:** AWS, MongoDB The team's own capstone documents — vision statement, both requirements documents, the development plan, the quarterly presentations and the final poster — are published alongside this project page and are the primary record. They are also available from the [UCSB capstone team page](https://capstone.cs.ucsb.edu/past22.html). These notes describe the project as documented. Availability of a public demo is separate from the project description.