Evaluating an AI‑Based Clinical Decision Support System in a Virtual Reality Intensive Care Unit

26.02.2026, Abschlussarbeiten, Bachelor- und Masterarbeiten

The Human-Centered Computing and Extended Reality Lab of the Professorship for Machine Intelligence in Orthopedics seeks applicants for Bachelor/Master Thesis for the Summer Semester 2026.

Abstract

This thesis will extend an existing virtual reality (VR) simulation of an intensive care unit (ICU) into a testbed for AI‑based clinical decision support systems. The student will integrate de‑identified into the VR monitors (e.g., heart rate, blood pressure, SpO₂) and connect an existing predictive model that runs in the background and raises alerts when a risk threshold is exceeded.

Background & Motivation

AI‑based decision support in the ICU depends on both reliable data streams (vitals, labs, scores) and robust integration of predictive models that continuously estimate patient risk. Evaluating these components directly in real ICUs is challenging because of safety constraints, technical complexity, and limited experimental control. A VR ICU “digital twin” provides a controllable environment to replay real patient trajectories, visualize monitoring data, and embed AI models that generate risk estimates and alerts in real time. While most existing work focuses on education or patient‑facing use, this thesis targets the technical integration layer: designing and validating a modular, message‑driven architecture that streams ICU time‑series data into VR, connects AI prediction services, and renders their outputs on virtual monitors with acceptable latency and fidelity, laying the groundwork for future studies on clinician interaction and trust.

Student’s Task

The primary objectives for the student working on this thesis will be:

• Design and implement a data pipeline to stream de-identified/synthetic ICU time‑series data (vital signs, key labs) into VR monitors and devices.
• Integrate an existing predictive model and implement logic to trigger in-VR alerts based on chosen threshold
• Design and implement in‑VR visualizations of model outputs (e.g., risk indicators, alert icons) and ensure they stay synchronized with the underlying data streams.
• Plan and conduct a user evaluation with domain experts to gather feedback and derive implications for future improvements

Technical Prerequisites

Students should have previous knowledge in developing 3D applications using the Unity 3D engine. Basic Python skills for handling data and integrating or wrapping an existing prediction model are beneficial. They should be motivated to prioritize usability and user experience and willing to collaborate with both novices and experts to enhance and test the system. German communication skills are beneficial, especially for coordinating and conducting feedback sessions with clinical staff, but are not mandatory.

Please send your transcript of records, CV and motivation to: Luisa Theelke (luisa.theelke@tum.de) with CC to hex-thesis.ortho@mh.tum.de

You can find more information and other topics for theses on our website: https://hex-lab.io

Literatur
Theelke, L., Metzler, F. L., Kreimeier, J., Hauer, C., Binder, J., & Roth, D. (2023, October). Investigating the Effects of Selective Information Presentation in Intensive Care Units Using Virtual Reality. In 2023 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) (pp. 195-204). IEEE.

Theelke, L., Beksultanow, D., Marquardt, L., Gulde, P., Vallines, L., & Roth, D. (2025, March). Advancing Critical Care Skills: Immersive VR Training Powered by Real-World Patient Data. In 2025 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW) (pp. 1660-1661). IEEE.

Kontakt: hex-thesis.ortho@mh.tum.de, luisa.theelke@tum.de