GPU-accelerated 3D multimodal visualization techniques for tumor resection in neurosurgery

Brain tumor resections, especially when the tumor is deeply embedded in the brain, are high-risk procedures. The difficulty lies not only in removal of the tumor itself but also in the process of gaining access to it without unnecessarily damaging surrounding tissues and brain structures. To highlight these tissues and structures different image modalities are needed such as CT for bone structures, MRI for brain matter, CT/MRI angiography for blood vessels, fMRI for cortical activation regions and diffusion tensor imaging (DTI) for fiber tracts. Furthermore, to correctly assess the spatial relation between these structures it is important to visualize and interact real time with the image data in 3D. For this purpose, we propose a new rendering algorithm based on Graphics Processing Unit (GPU)-accelerated raycasting and depth peeling to visualize multiple volumetric datasets intersected with an arbitrary number of opaque or semi-transparent geometric models. Such models may represent foreign objects relevant for surgical applications such as virtual surgical tools, 3D pointers, measuring tools or grid lines for spatial orientation.

@InProceedings { BVPT08,
  author       = "Brecheisen, R. and Vilanova, Anna and Platel, B. and Haar Romeny, B.M. ter",
  title        = "GPU-accelerated 3D multimodal visualization techniques for tumor resection in neurosurgery",
  booktitle    = "CARS Conference Proceedings",
  year         = "2008",
  url          = "http://graphics.tudelft.nl/Publications-new/2008/BVPT08"
}