Peter - for some reason my mail system won't allow me to reply to
[log in to unmask]
Could you forward this for me, please.
I have tried to use mcflirt for exactly this same reason, in data from a
tracking task where we know the ocular target position and know the
subjects will have been tracking that with eye movement.
I first choose a very small volume around one eye, from data that had
been mcflirted to remove overall head motion, but got poor mcflirt
output of this "eye movement" data. I got better results by masking the
eye within a larger volume - 10x20x5 voxels within 30x30x10 I think, as
the flirt seems to need moderately sized volumes. Mcflirt seemed to work
fine on this data, but I am not convinced that the motion output
resemble anything like real eye movement.
If you look carefully at figure 2B of Tregallas et al, you will see no
noise at all on the reconstructed eye position data. In my hands, the
correlation between "eye position" and target position after mcflirt was
very poor. I'm not sure why they get such enormous effects of eye
movement in their data, but it seems that the FMRIB scanner doesn't have
a strong eye movement signal, unfortunately.
--
Regards,
Chris
_____________________________________________________________
Chris Miall Tel: +44 1865 282162
Professor of Neuroscience Fax: +44 1865 272469
University Laboratory of Physiology Mobile: 07815 296483
Parks Road [log in to unmask]
Oxford OX1 3PT, U.K. http://www.physiol.ox.ac.uk/~rcm
=============================================================
Arun Bokde wrote:
>
> Hello FSLers,
>
> A recent paper (Tregellas J et al, HBM, vol 17: 237-243) shows that a
> measure of eye movement can be obtained by using the standard EPI
> sequences. I am interested in trying it out using FLIRT, as the method
> described basically is a rigid body registration process on the time
> series of each eye. Is there any reason that FLIRT would have problems
> with a time series file that has perhaps 10 x 10 voxels per slice and
> 4-5 slices ? Is there a program that could produce a 4D avw file from
> another 4D avw file based on an ROI ? I was looking at avwroi but the
> help info is not very clear to me.
>
> TIA
>
> Cheers, Arun
|