Thanks for the explanation.

So now the steps would be:
1. Normalization by the maximum intensity in an angio dataset:
$ 3dcalc -a Angio+orig -expr 'a/max_intensity_Angio' -fscale -prefix Angio_norm

2. Resampling the angio dataset to lower resolution:
$ 3dfractionize -template func+orig -input Angio_norm+orig -prefix Angio_frac -preserve

It seems to me that 3dfractionize here (without -clip and -wrap options) is doing almost the same thing as 3dresample if I use the command below
$ 3dresample -master func+orig -rmode Li -prefix Angio_resamp -inset Angio_norm+orig
So what's the difference on the usage?

3. Detect the blood vessels in the angio dataset
$ 3dmerge -1noneg -1clip 0.2 -dxyz=1 -1clust 1.01 8 -prefix Angio_vas Angio_frac+orig

Advice and comments will be greatly appreciated.

Lydia