Dear Steve,
when you look at the images do you think that at least the RF bias field is
corrected sufficiently? In the other words, provided that the images cover
"enough" brain, would normalized mutual information (eventually other) cost
function work on these images?
Is there any other approach (though not implemented in FSL) to register these
type of images, which in your oppinion could work. I know, my questions depart
from FSL, nevertheless, I would appreciate your expert opinion?
Best wishes,
Martin
Quoting Stephen Smith <[log in to unmask]>:
> Hi - the idea of using the input image here as it's own weighting mask
> is
> a good one, given that it's outline has gotten seriously corrupted. I
> would erode the mask a couple of times though to really get rid of
> edge
> effects (using -ero in avwmaths).
>
> HOWEVER, this particular input image really is totally screwed and
> even
> the above probably can't be made to work; whatever preprocessing was
> carried out removed huge amounts of brain, so you don't stand much
> chance
> of registering to the reference. Really, you need to go back and fix
> the
> preprocessing!
>
> Cheers, Steve.
>
>
>
> On Fri, 29 Oct 2004, Martin Kavec wrote:
>
> > Hello group,
> >
> > I have no luck to register a high resolution T1-weighted (1.5T)
> volume
> > (stored at http://www.uku.fi/~kavec/reference_brain.tar.gz ) and
> > T1-weighted (0.12T) volume (stored at
> > http://www.uku.fi/~kavec/input_brain_bias.tar.gz ).
> >
> > Problematic is, I believe, the input volume.
> > (1) It was masked by the manufacturer of the MR system with an
> ellipsoid
> > mask to hide artefacts due to B0 inhomogeneities. The mask
> > introduces false boundaries in the image, which confuses
> > registration.
> > (2) Additionally, the images were acquired with a
> > large diameter surface coil producing large intensity variations,
> which
> > I did my best to correct using FAST.
> >
> > To address (1), I am created a brain mask (stored at
> > http://www.uku.fi/~kavec/input_brain_mask.tar.gz ) using BET and I
> use
> > this as an input weighting volume.
> >
> > I would expect that normalize mutual information should be able to
> > register the imaged using DOF=6-9, but after failing in this I tried
> > virtually every combination of DOF<6 and cost function, but as I
> mentioned
> > without any good results.
> >
> > I would appreciate any input as to what am I doing wrong. Many thanks
> in
> > advance.
> >
> > Martin
> >
>
> --
> Stephen M. Smith DPhil
> Associate Director, FMRIB and Analysis Research Coordinator
>
> Oxford University Centre for Functional MRI of the Brain
> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
>
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>
|