thanks for your advice! No, I'm running Linux. I'm currently trying to find
a parameter set that allows realignment of multi-frame F-DOPA scans.
Unfortunately, the results I've obtained so far are not convincing. I've
compared them to SPM's results and there is some motion that is quite
visible when I scroll through the time series using FSLview, which is not
present after realignment with SPM.
Now, I'm trying to find out how to opimize the whole process, so I can get
the same results using mcflirt. I've tried all of the available cost
functions (except for normmi), setting -smooth 0.0 (in my DOPA frames, the
scalp shows up nicely, but is naturally closer to the border of the FOV than
the other brain structures, so I thought deweighting peripheral voxels might
hinder correct registration), and setting -stages 4. Unfortunately, these
changes did not produce satisfactory results.
Do you have any further suggestions how I should proceed? I understand that
PET frames are quite different from an fMRI time series, but I hope there's
a way to make mcflirt work for this.
> -----Ursprüngliche Nachricht-----
> Von: FSL - FMRIB's Software Library
> [mailto:[log in to unmask]] Im Auftrag von Mark Jenkinson
> Gesendet: Mittwoch, 5. Mai 2004 09:10
> An: [log in to unmask]
> Betreff: Re: [FSL] mcflirt exception
> Are you using cygwin?
> It may be a problem with that binary - we could not replicate
> the previous problem that Thomas had.
> However, I think it went away for Thomas if he stopped using
> the -meanvol option. In general we do not recommend using
> the -meanvol as we have done tests (see Jenkinson, M.,
> Bannister, P., Brady, J. M. and Smith, S. M., Improved
> Optimisation for the Robust and Accurate Linear Registration
> and Motion Correction of Brain Images/, NeuroImage/, 17(2),
> 825-841, 2002.) and meanvol doesn't turn out to be the best.
> It isn't far off but it does degrade the results. Partly
> this is because the initial mean is not made from well
> registered volumes, but partly it is because the mean is
> inevitably smoother than an individual volume which means
> that it is a better match if the registration smooths the
> original volume by adding some offset and making the
> interpolation do some smoothing. This is why we use a single
> volume reference in FEAT.
> So, try it without the meanvol and see if it works.
> All the best,
> On Tuesday, May 4, 2004, at 02:26 pm, Johannes Klein wrote:
> > Hi everybody,
> > I've come across a problem that is similar to the one
> Thomas Mierdorf
> > had posted previously. I'm trying to realign a series of PET frames
> > using mcflirt and the -meanvol and -plots option, which
> results in an
> > "Image Exception : #5 :: Out of Bounds (time index)" error in the
> > final iteration
> > of the second, realign-to-mean process.
> > The command line is:
> > mcflirt -report -plots -meanvol -in filename -out filename_mcf
> > If I don't use the -plots option, everything works, but I'd really
> > like to
> > be able to check how much patient movement there is in my
> time series.
> > It seems like there's a problem with the generation of the
> > transformation
> > parameters file.
> > The output of avwhd, avwstats -r -R can be found at
> > http://www.mpifnf.de/~johannes/mcflirterror.txt
> > I'd appreciate any help with this!
> > Thanks
> > Johannes