Dear All, I tried to extract a surface from a MRI scan (T1, 1.5T) using Freesurfer 3.0.3 on a Linux computer. I just run "recon-all" script for the same scan 6 times (same computer, no human interaction). To my supprise, I got 6 different results ( different #triangles, #vertices). Go back to check the intermediate results, all T1.mgz, WM.mgz are identical, but lh.origdifferent. Can someone tell me why my results are not consistent? Thanks very much for your help. Li-Chen
Li-Chen,
Some of the utilities make use of random number generation in their algorithms, specifically: mris_smooth, mris_sphere, mris_topo_fixer and mris_ca_label. While the number of vertices can change because of this, the resulting stats (cortical thickness) should not differ by more than 0.5%.
If you need results to be perfectly consistent for a subject from run to run, such as the case for our nightly testing scheme on a reference subject, then you can add the -norandomness flag to recon-all, which will seed the random number generator with the same number (instead of seeding from the current time).
Nick
On Tue, 2006-12-05 at 12:16 -0600, Lichen Liang wrote:
Dear All, I tried to extract a surface from a MRI scan (T1, 1.5T) using Freesurfer 3.0.3 on a Linux computer. I just run "recon-all" script for the same scan 6 times (same computer, no human interaction). To my supprise, I got 6 different results ( different #triangles, #vertices). Go back to check the intermediate results, all T1.mgz, WM.mgz are identical, but lh.orig different. Can someone tell me why my results are not consistent? Thanks very much for your help. Li-Chen _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Li-Chen,
there is some intentional randomness in the algorithms. Try overlaying the surfaces and you'll see that they are visually identical I would think. Nick: is there a switch to recon-all to specify what the seed should be for the random # generators?
thanks, Bruce
On Tue, 5 Dec 2006, Lichen Liang wrote:
Dear All, I tried to extract a surface from a MRI scan (T1, 1.5T) using Freesurfer 3.0.3 on a Linux computer. I just run "recon-all" script for the same scan 6 times (same computer, no human interaction). To my supprise, I got 6 different results ( different #triangles, #vertices). Go back to check the intermediate results, all T1.mgz, WM.mgz are identical, but lh.origdifferent. Can someone tell me why my results are not consistent? Thanks very much for your help. Li-Chen
On Tue, 2006-12-05 at 13:31 -0500, Bruce Fischl wrote:
Hi Li-Chen,
there is some intentional randomness in the algorithms. Try overlaying the surfaces and you'll see that they are visually identical I would think. Nick: is there a switch to recon-all to specify what the seed should be for the random # generators?
The -norandomness flag can be added to recon-all. It does not take a number as input. It will seed with the number '1234', which, by having this number hard-coded in recon-all, means that you don't have to remember what seed was used on some prior run.
Nick
thanks, Bruce
On Tue, 5 Dec 2006, Lichen Liang wrote:
Dear All, I tried to extract a surface from a MRI scan (T1, 1.5T) using Freesurfer 3.0.3 on a Linux computer. I just run "recon-all" script for the same scan 6 times (same computer, no human interaction). To my supprise, I got 6 different results ( different #triangles, #vertices). Go back to check the intermediate results, all T1.mgz, WM.mgz are identical, but lh.origdifferent. Can someone tell me why my results are not consistent? Thanks very much for your help. Li-Chen
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Lichen Liang -
Nick Schmansky