Dear Martin,
> it was not possible for me to keep my patients in the magnet for
> extended
> periods that's why I am also left with approach discussed in this
> thread.
my warning was not meant as a criticism in any way, and I do
understand that time can sometimes be more of an issue than I am used
to. It was more intended to discourage people from thinking there is
no need to acquire fieldmaps/blip-up-down data. And at the end of the
day the time involved in getting a fieldmap or one additional EPI with
reversed flip polarity is quite short and acceptable for most
applications.
> Disregarding the important issues discussed later in this thread I
> wonder,
> whether you haven't tried segmenting WM from T1, skeletonize both
> the FA nad
> T1-WM and FNIRT them. This way you would work on the centers of the
> track in
> the both images and may avoid the widening.
I am not aware of anyone attempting this, and I wouldn't expect it to
work particularly well. First of all it hinges on a skeletonized
version of the white matter segment to be identical to the FA derived
skeleton. Which I don't think it will. Secondly voxel based non-linear
registration tends to work poorly on sparse data, i.e. data which is
only defined for a small subset of the volume.
Jesper
|