I can exclude a cylindrical inner region from the density distribution easily enough, I was just wondering whether this does what you want or whether the whole ray computation needs to terminate when hitting the boundary.
I guess you really need the latter because of transparency/opacity issues...