You could try the 3dIntracranial program. It works about 80-90% of the time, at least on T1-weighted images (MP-RAGE and SPGR styles of acquisition). I don't know how well it would work on angio data. If you have a T1-weighted structural, you could try 3dIntracranial on it; if that works, then you could use that as a mask (e.g., via 3dAutomask and 3dresample and 3dcalc) to mask off angio datasets.
I take it that you want to do projections (via the Render plugin) on the angio data?
bob cox