State-of-the-art GPU volume rendering

Question

what is the current technique for implementing a GPU based volume renderer in the medical context, ie reasonable amounts of image data (512x512x1500 or more at 16bit) and interactive/real-time performance with optimal image quality? Is it still the raycasting approach introduced by Krüger and Westermann (color cube etc.)? Or have CUDA/OpenCL changed the traditional Cg/GLSL game?

Thanks for any hints, paper references etc.! Matthias

Answer 1

Nothing has really changed since then. There are different versions of CUDA/OpenCL/DirectCompute implementations floating around, but they are basically do the same thing.

Here's nice SIGGRAPH 2010 course Volumetric Methods in Visual Effects which summarize common techniques.