John,
I'm thinking more about a self-test (like the FSL FEEDS package) to
verify that the local admins got the matlab install right and there
are no surprises from libraries or configs outside of the spm code.
The FEEDS script performs an automated series of FSL analyses on known
input and then quantitatively compares the local results to known-good
output. This verifies that there are no missing software dependencies
or version mismatches that might 1) cause fatal errors, or, even
worse, 2) produce aberrant results that aren't easily detectable as
such. Not all software errors produce grossly deformed output. :)
This is especially comforting when spreading analyses across multiple platforms.
warm regards,
Stephen
On Thu, Apr 3, 2008 at 8:14 AM, John Ashburner <[log in to unmask]> wrote:
> If the images to segment are not initially in good alignment, then the
> expected results are pretty bad. In almost all cases, if there are problems,
> they are due to the initial affine registration going wrong.
>
> Best regards,
> -John
>
>
>
> On Thursday 03 April 2008 11:55, Stephen Towler wrote:
> > Hi all. Is there a validation dataset and/or process to verify that an
> > SPM5 installation on a new platform is performing as the authors
> > intend? I'm particularly interested in verifying that update_1111
> > tissue-type segmentation (T1) and VBM are producing expected results.
> >
> > Thanks in advance,
> > Stephen
>
>
--
Stephen Towler
[log in to unmask]
352-294-0048 office
352-682-5231 mobile (*NEW*)
352-392-8347 fax
Leonard Lab
Department of Neuroscience
PO Box 100244
University of Florida HSC
Gainesville, FL 32610
|