Hi,
The suggested value of 0.5mm for smoothing was based on a test performed
for a given set of EPI data, collected on our 3T machine here. The results
can be found at:
http://www.fmrib.ox.ac.uk/~mark/work/hbm2001.ps
However, you can see from the results that it doesn't make a lot of
difference.
Hence, on a different scanner, at a different field strength and
*especially*
for different phase-encode directions, the results could be better with a
different smoothing, possibly even up to 5mm as you've used.
What the smoothing controls is the amount of regularisation (3D Gaussian
smoothing
in this case) that is applied *after* the unwrapping takes place. Hence
this tries
to increase the SNR in the field map values, and suppress local outliers
to some
extent. So the optimal smoothing/regularisation will also depend on your
field map sequence and its inherent SNR. The better it is, the less
smoothing
you should need.
So at the moment I'd recommend going with what you find works empirically.
However, it would obviously be nicer if it were a bit more quantitative
and we
are working on tests and new methods to help in this way.
All the best,
Mark
Johanna Pekkola wrote:
>Hi!
>
>I`m planning to unwarp my epi-data using fieldmaps and Fugue and I would
>like to ask about the -s (smoothing) option in Fugue. The default represented
>in FSL web page was 0.5 (mm).
>
>I tried with a few different values ans 0.5 did not seem to be different from
>the unsmoothed unwarped image. Instead, -s of 5 (mm) seemed to my eye "better",
>so, resembling significantly more the T1 image in shape.
>
>So - what is the -s kernel applied to in the unwarping process? Is it
>smoothing the phase "jumps" in the fieldmaps? And could the sigma of the
>kernel required depend on the field strenght?
>
>Thanks in advance,
>Johanna Pekkola
>TKK/LCE
>Finland
>
>
|