Bambusman is the expert here, but making a volumetric M42 for example is *really* hard. but yeah nebulae having clearly a rotation or reflection plane symmetry could be the first ones to be emulated with volumetric techniques. An hybrid approach is more probable in the near future, as in using both volumetric and raster methods for the same object, but it's still work in progress and, if ever, it needs huge optimization.
About the 3d models from those agencies, they use huge amounts of computational power to calculate those numerical data, while volumetric models need something like a parametric 3d function to approximate the real shape for our poor computers.
There are techniques involving fractal algorithms in order to match somehow real images with mathematical functions, used mainly to search new high quality audio-video compression mechanisms. Add a third dimension and I guess we are talking about frontier informatics