Projections and MR Angiography
The direction of projection is rotated around, and the result is a series of 2D images that show in each pixel the brightest (or darkest) voxel along the projection ray. The vessels stand out as 1D tubular structures in this type of animation. It is a very useful way to look at MRA data.
This technique could be used as a way of detecting vessels automatically as well. The idea is to do many such projections from a lot of directions covering the ½-sphere, and for each voxel, keep track of from how many directions it is the max (or min) along a projection. Voxels that are quite often the extreme value are likely to be in a vessel. Then some sort of connectivity criterion can be used to exclude isolas and/or spackle over gaps.
Each projection, on a 256x256x128 volume, should take less than 0.5 s, so in a couple minutes the basic process could be carried out. The 'extract_' functions in AFNI's cox_render.c could be modified to carry out this computation relatively easily. Someday ....
Projections and MR angiography