I am running Linux (redhat 9) so i'll have a go with the slower flirt.
the registrations were fine for the other subject for which i completed
a higher level analysis. the runs were quite variable though - i could see that
from eye, looking at the locations of the activations (but each run
was only made up of 30 volumes). i seem to have a
problem looking at time series though - for most functional files that
i load into fsl view, when i click time series nothing appears - even for
stats images from the higher level analysis. the only
time i am able to see time series is when i load an orignial series_2.*.hdr
and overlay a stats image onto to it. is this correct?
In message <[log in to unmask]> FSL - FMRIB's Software Library <[log in to unmask]> writes:
> Hi Jane,
> I'm afraid you are experiencing the effects of an intermittent
> bug in flirt which we think is caused by the compilers we use,
> which was unexpected. What machine are you running it on?
> If it is linux (redhat 8 or 9) then a non-optimised version of
> flirt, which is slower but doesn't crash, can be installed.
> You might need to get your sysadmin to do this for you.
> If you are not running on linux/redhat 8 or 9 then let us know.
> As for the group analysis, this can happen if there is a lot of
> variation between the runs. Does the data from different runs
> appear significantly different in fslview (look at the timeseries)?
> Also, if your registrations aren't working then you won't get
> any good higher level results. Do all the registrations in the
> registrations reports look good?
> If you can't see anything obviously wrong in the above, then let
> us know and we can try to have a look at your data/design.
> All the best,
> Jane Aspell wrote:
> >I'm new to FSL and am having a few problems. When I run a first level
> >analysis using FEAT for one of my subjects something seems to go wrong with
> >the co-registration of the high res image to the standard brain. In Feat
> >watcher i get the following error:
> >/usr/local/fsl/bin/convert_xfm -matonly -inverse -omat standard2highres.mat
> >Could not open matrix file highres2standard.mat
> >Cannot read input-matrix
> >/usr/local/fsl/bin/slicer highres2standard standard -s 1 -x 0.35 sla -x 0.45
> >slb -x 0.55 slc -x 0.65 sld -y 0.35 sle -y 0.45 slf -y 0.55 slg -y 0.65 slh
> >-z 0.35 sli -z 0.45 slj -z 0.55 slk -z 0.65 sll ; /usr/local/fsl/bin/convert
> >-colors 100 +append sla slb slc sld sle slf slg slh sli slj slk sll
> >highres2standard.gif ; /bin/rm -f sla slb slc sld sle slf slg slh sli slj
> >slk sll
> >Cannot open volume highres2standard.hdr for reading!
> >convertb: no delegate for this image format (sla).
> >convertb: no delegate for this image format (slb).
> >convertb: no delegate for this image format (slc).
> >convertb: no delegate for this image format (sld).
> >convertb: no delegate for this image format (sle).
> >convertb: no delegate for this image format (slf).
> >convertb: no delegate for this image format (slg).
> >convertb: no delegate for this image format (slh).
> >convertb: no delegate for this image format (sli).
> >convertb: no delegate for this image format (slj).
> >convertb: no delegate for this image format (slk).
> >convertb: no delegate for this image format (sll).
> >convertb: Missing an image file name.
> >/bin/rm -f highres2standard.hdr highres2standard.img
> >And I found that the highres2standard.mat transform is missing from the /reg
> >directory in the relevant .feat directory.
> >So i tried to create it with the command:
> >flirt -in highres -ref standard -omat highres2standard.mat
> >But that just returns 'Segmentation fault'
> >I haven't had this problem with my other subject. Any ideas?
> >My second problem is with higher level analysis. I ran successful first
> >level analyses for the 5 series(runs) for a single subject and got good
> >activations for my contrasts (z~6) but when i run the higher level analysis
> >to put all the runs (which are simply repeats of the same conditions)
> >together all the activation seems to disappear! I re-ran the higher level
> >analysis with a lower z threshold (1.0) and was then able to see activations
> >(z scores <3 could be seen). Is this the correct thing to do? I thought
> >the activations would be more significant, not less, when I average all
> >my runs together so I'm puzzled.
> >Thanks very much for your help.
> >Jane Aspell