yes, that was our goal some time ago and we thought we had succeeded in
tracking down and fixing any randomness if the same seed is specified
Bruce
On Mon, 23 Jun 2014, Harms, Michael wrote:
>
> Actually, if you run 'recon-all' a second time (from scratch) on the same
> subject, on the same pl
Actually, if you run 'recon-all' a second time (from scratch) on the same
subject, on the same platform/architecture, and you use the default
recon-all settings (i.e., don't use the -randomness flag, or specify your
random seed via the -rng-seed flag), then you should get identical
results, becaus
There of course is some variability in the analysis due to the way the
segmentation happens. Your results are certainly not out of the ordinary for
the volume you are viewing. On average you can expect a ~1% variance between
several scans of the same subject.
If you want to know more I certainl
Wang,
what command-line flags were included? can you send me the recon-all.log file
for these runs? The results should be the same with repeated runs if the
default flags were used.
Nick
On Jun 21, 2014, at 10:20 PM, wangkangcheng_gmail
wrote:
> Dear Bruce
>
> Yes. The following is an
Dear Bruce
Yes. The following is an example result for the subject (The id from Sub_1 to
Sub_10 is the same subject) that I calculate ten times.
rh.aparc.a2009s.volume rh_G_and_S_frontomargin_volume
Sub_1 1713
Sub_2 1950
Sub_3 1789
Sub_4 1783
Sub_5 1879
Sub_6 1778
Sub_7 1892
Hi Wang
what do you mean "the number of each index"? And do you mean you analyzed
the exact same data 10 times on the exact same hardware?
cheers
Bruce
On Sat, 21 Jun
2014, wangkangcheng_gmail wrote:
> Dear experts
>
> I used freesurfer (Version 5.3) to calculate one subject for ten times and
