Ooops - I pressed send accidentally on the last email before I was finished.
What I was going to add was that your images also had a lot of extraneous
material near the edge of the brain, so I would definitively recommend creating
a masked version. Unfortunately you can't use BET easily because you have
such a restrictive FOV and BET is designed to work on whole brain images.
Also, for a small number of slices like this it is
quite reasonable to use 2D mode, which is also a lot faster. To do all of
this use the command line:
prelude -a abs_masked -p phs_in_rad -u result -s
where I generated phs_in_rad by:
avwmaths phs -div 1289 -sub 25.3 -min 3.14 -max -3.14 phs_in_rad
Note that this also clamps the few voxels outside the -pi,pi range so
that they then lie within it again.
In general though the only problem was that prelude expects a phase image
to be in radians, between -pi and pi. So, rescaling and removing the
appropriate mean (as above) should fix it.
If you could find out exactly what the numbers in the phase image were
meant to refer to, then you should be able to do even better by getting
the rescaling/demeaning absolutely correct.
All the best,
Mark
Martin Kavec wrote:
>
> Hello all lucky FSL users,
>
> after being silent for a while, I started to play aroud with my data and try
> to improve the quality with amazing FSL.
> I need to unwrap the phase image taken with long TE SPGR. I simply ran
> prelude -a abs.hdr -p phs.hdr -u result -f
> Since the result was what I would like to have I first did BET on abs.*. The
> situation slightly improved, but phases were unwrapped. Actually, phase image
> looked only like a masked original and phases were untouched.
> I believe, I am again doing something wrong, since I have had only
> extraordinary results with FSL so far.
> I am just uploading the source images to fsl site (or actually to Stephen
> Smith e-mail). One more note concearning the phase image. It becomes from GE
> Signa CVi scaner (software version 8.4) and I am not completely sure, how to
> scale it down to prelude recquirements. I just devide each pixel by 1000 on
> this data set
>
> I would be greateful, if I can get any help.
>
> All the best to all of you.
>
> Martin
|