VISTA-MED: integrating AI-based MRI segmentation with virtual reality for medical education

Abstract

Innovations in Artificial Intelligence (AI) and Virtual Reality (VR) are increasingly applied in medical imaging and education by enabling automated segmentation and immersive visualization. Traditional anatomy and imaging instruction often provide limited spatial interactivity, limiting comprehension and diagnostic reasoning. This study introduces VISTA-MED, an integrated AI-VR platform designed to enable interactive exploration through real-time MRI visualization and AI integrated segmentation. The system employs the nnU-Net architecture for automated segmentation of brain tumor regions using the BRATS2020 dataset. Preprocessing included bias field correction, image registration to the MNI152 template, and noise reduction. Segmentation outputs were converted into 3D-compatible formats and integrated into a Unity-based VR environment to allow interactive exploration of anatomical structures. Performance was evaluated using Dice coefficient, and precision-recall analysis. The AI segmentation model achieved a mean Dice score of 0.917, indicating accurate tumor boundary delineation. Segmentation outputs were evaluated through overlay comparison with original MRI volumes, showing alignment with anatomical structures. Internal testing of the VR environment confirmed smooth rendering and stable frame rates, enabling real-time interaction with the 3D brain model. VISTA-MED demonstrates the technical feasibility of combining AI-driven segmentation with immersive VR visualization for interactive exploration of MRI data. The prototype establishes a foundation for future development of educational modules and provides insights into design considerations for AI-VR systems in medical imaging. Future work will focus on usability studies, scalability, and educational impact assessment to support broader adoption in clinical and academic settings.

Ähnliche Arbeiten