some things to consider.
if you perform any realignment steps on your data (including
eddy-current correction which is the usual first step) the data will be
smoothed by the image interpolation.
So TBSS actually leads to more smoothing because of the non-linear
coregistrations performed on the DT images.
as Matt says, it's worth checking your sequence. you should be
collecting at least one B0 image for every ~6 DWIs. that will help with
SNR. I know the standard Siemen's sequence and potentially Phillips also
don't allow you to collect mulitple B0 images by default. You images
resemble those from the standard Siemen's sequence.
cheers
s
Matt Glasser wrote:
> It depends on what you plan to use the data for. This will create big
> problems for tractography, but might be acceptable for TBSS or ROI based FA
> analyses. You will need to keep in mind that smoothing like this will
> increase partial volume effects significantly. It is partly for this reason
> that TBSS was developed, as smoothing has unpredictable effects on what is
> found to be significant in VBM based analyses of FA data (where smoothing is
> absolutely required), see Smith et al 2006 in Neuroimage for more details on
> this issue.
>
> If it is still possible to improve your acquisition methods, it is worth
> revisiting why you are getting this much noise in your FA images, as there
> might be ways to fix this problem.
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
> Of Jiansong Xu
> Sent: Saturday, December 19, 2009 12:02 PM
> To: [log in to unmask]
> Subject: [FSL] DTI: smooth or not smooth
>
> Dear All:
>
> I would like to know if it is a good idea to smooth DTI raw images
> before creating FA and other maps.
>
> The first attached FA map was created without smoothing the raw images
> and it is very "noisy". The second attached FA map was created after
> smoothing the raw images by using: fslmaths input -s 3 output.
> Relative to the first FA map, the second map appears much more "smooth".
>
--
========================
Scott Kolbe
Neuroimaging Group
Florey Neuroscience Institutes and
Centre for Neuroscience
University of Melbourne
VIC, Australia, 3010.
ph: +61 3 8344 1929
email: [log in to unmask]
website: www.neuroimaging.org.au/index.php?id=383
|