Hi Live,
> Hi Gwenaelle,
>
> Thank you very much for your answer. I checked through my
> individuals,
> and there are many of the individuals that contributes to
> reducing the
> mask, unfortunately.
>
> From your answer below from 2009, you say that Steve
> changed the way to
> create the mask in the latest version of FSL. How was the
> mask created
> in the older versions, and in what version was that? And
> how do you
> think the mask will be if I try an older version of fsl?
I'm afraid this hasn't been changed yet... But the idea is that, instead of using the intersection mask to directly mask the all_FA, you would use it instead in randomise. That way, you would not loose the information on the all_FA(_skeletonised) and would be able to identify the problematic subjects (like what you've just done), but would still be unable to get any results in a region where information is missing.
So for the moment unfortunately, your results will only be valid where all the FA images overlap, as indicated by the mask.
Hope this makes sense,
Gwenaelle
> Gwenaëlle DOUAUD wrote:
> > Hi Live,
> >
> > yes, I suspect this is the same problem that discussed
> below... It might be easier for you to spot "missing" voxels
> if you re-create the unmasked all_FA image by running:
> >
> > ${FSLDIR}/bin/fslmerge -t ../stats/all_FA
> `$FSLDIR/bin/imglob *_FA_to_target.*`
> >
> > in your FA directory.
> >
> > If you look in fslview at the timeseries by pointing
> at voxels out of the mask, you should be able to spot
> quickly which images are causing the problems.
> >
> > Your #n is very large, so if these are only fews scans
> reducing massively the size of the mask, I would remove them
> from the analysis. As Steve mentioned, there might be at
> some point some ways of running outlier detection in tbss,
> but not yet... Thus, if a non negligible number of scans are
> contributing to reducing the mask, then you might
> unfortunately need to only look in the white matter covered
> by this mask.
> >
> > Hope this helps,
> > Gwenaelle
> >
> >
> >
> >
> >> De: Live Eikenes <[log in to unmask]>
> >> Objet: [FSL] tbss_3_postreg mean_FA_mask
> incorrect
> >> À: [log in to unmask]
> >> Date: Vendredi 11 février 2011, 9h27
> >> Dear FSL list,
> >>
> >> I am running tbss on 825 individuals (age between
> 50 and
> >> 65), and I have
> >> got as far as tbss_3_postreg without problems or
> any error
> >> messages. I
> >> am using the tbss_2_reg -T option with the
> FMRIB58_FA as
> >> the target, and
> >> the tbss_3_postreg -S option.
> >>
> >> But when I looked at the results in the stats file
> from
> >> tbss_3_postreg,
> >> my mean_FA_mask coveres a small area of the brain
> (it
> >> covers the corpus
> >> callosum and almost the whole brain from inferior
> to
> >> superior, but much
> >> the right and left sides of the brain is missing).
> And
> >> since the
> >> mean_FA_mask is used for the all_FA, mean_FA and
> >> mean_FA_skeleton, these
> >> also covers a small part of the brain.
> >>
> >> I searched the FSL archives, and found the
> following answer
> >> to what
> >> seems to be the same problem:
> >>
> >>
> >> "On Thu, May 21, 2009 at 12:05 PM, Gwenaëlle
> DOUAUD
> >> <[log in to unmask]
> >> <https://www.jiscmail.ac.uk/cgi-bin/webadmin?LOGON=A3%3Dind0905%26L%3DFSL%26E%3Dquoted-printable%26P%3D2234062%26B%3D--0016e645fa8a7fec87046a6ea584%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D>>
> >>
> >> wrote:
> >> Hi again Sushravya,
> >>
> >> I think I know what's going on: in the most recent
> version
> >> of TBSS,
> >> Steve has changed the way to create the mask for
> the
> >> analyis. Basically,
> >> it now takes the *intersection* of all your FA
> maps.
> >> This is a good way to be aware of some data
> truncated and
> >> lack of
> >> overlap across subjects etc...
> >>
> >> So it might be the case that your mean_FA_mask is
> actually
> >> correct and
> >> points out as one image (possibly several?) being
> >> truncated... What
> >> happens if you load your all_FA in FSLview and
> turns on the
> >> movie loop?
> >> You should be able to spot the images contributing
> to the
> >> reduced
> >> mean_FA_mask and either remove them or try to fix
> what is
> >> wrong with
> >> them if possible.
> >>
> >> Cheers,
> >> Gwenaelle"
> >>
> >> So based on this does this mean that some of my
> images are
> >> being
> >> "truncated", and that I need to find these images?
> Before I
> >> ran the tbss
> >> analysis, I looked through all the FA images in
> all the
> >> individuals, and
> >> to me it looked fine.
> >>
> >> Do you have a solution to my problem?
> >>
> >> Best regards,
> >> Live Eikenes.
> >>
> >
> >
> --------------------------------------------------------------------
> >
> > Gwenaëlle Douaud, PhD
> >
> > FMRIB Centre, University of Oxford
> > John Radcliffe Hospital, Headington OX3 9DU
> Oxford UK
> >
> > Tel: +44 (0) 1865 222 523 Fax: +44 (0) 1865 222
> 717
> >
> > www.fmrib.ox.ac.uk/~douaud
> >
> >
> --------------------------------------------------------------------
> >
> >
> >
> >
>
