Different OSs will often have different math libraries which can generate different results. Usually, these are very small, but we still recommend that all group analyses use the same platform.
On 2/25/19 5:18 AM, Tim Schäfer wrote:
External Email - Use CautionDear Freesurfer experts,
we recently got access to a Linux-based cluster computer, and I need to run FreeSurfer pre-processing for several 100 subjects soon. We also have a Mac Pro with many cores that we used in the past and that could take part in the pre-processing.
I found an older reference which suggests you should not run the pre-processing for one study on different architectures/operating systems [1] because this may affect the results. Does this still apply to Freesurfer 6? What are your recommendations in this regard?
Also, out of curiosity: it is obvious why one should not use different Freesurfer versions, but what are the reasons for the reported differences between architectures/workstations?
[1]https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0038234
Tim
-- Dr. Tim Schäfer Postdoc Computational Neuroimaging Department of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy University Hospital Frankfurt, Goethe University Frankfurt am Main, Germany
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer