Dear FSL users,
We have just found a bug that affects vertex analyses run
with FSL4.1.4 (22 May 2009). It does not affect the FIRST
segmentations or volume measurements - only vertex analyses.
It affected the way that the groups are generated by concat_bvars.
There is a patch to concat_bvars which fixes it at:
http://www.fmrib.ox.ac.uk/fsldownloads/patches/
just download concat_bvars and copy it into $FSLDIR/bin/
This bug caused concat_bvars to produce an alphabetically ordered list
of files
internally. If you used an alphabetically ordered list (e.g. single
wildmask)
then there is *no* problem - which is why we haven't noticed before.
Also, if
you had equal group sizes with consistent naming then all stats are
likely to be
fine, but vectors and surfaces may be reversed. Otherwise, things
could be
quite different and you will need to re-run.
I would generally recommend re-running the patched version of
concat_bvars
to create a new file as it only takes seconds. Then check if it is
different from
the old version (check with: diff file1 file2). If it is different
then you should re-run
any vertex analysis that used it.
Many apologies for this bug.
I hope that it does not cause major problems for anyone.
Sorry,
Mark
|