Hi,
Sorry to dredge up an old thread, I was curious about the details on where randomness is introduced in the following programs (beyond the normal floating points issues):
mris_smooth -seed 1234 mris_sphere -seed 1234 mris_fix_topology -seed 1234 mris_ca_label -seed 1234
I assume that the Segonne, Grimson, and Fischl paper on the genetic algorithm for cortical surface correction applies to mris_fix_topology? Is it also the case that the randomness is injected in other min/maximization-type algorithms in the other functions?
RE: [Freesurfer] variability Nick Schmansky Thu, 21 Dec 2006 10:14:12 -0800
Valentina,
The randomization in the algorithms accounts for about a 0.02% change in thickness results. You can disable randomness (or rather, force seeding the rng with same number) by including the flag -norandomness with recon-all. This flag is available in v3.0.4.
In older versions, do this to disable randomness: create a file called xopts containing this text:
mris_smooth -seed 1234 mris_sphere -seed 1234 mris_fix_topology -seed 1234 mris_ca_label -seed 1234