Hi,
another option may be this:
http://projecteuclid.org/euclid.ba/1440594946
In short, we have implemented a spatially adaptive, conditionally
autoregressive model: The data itself is not smoothed. Instead, we
adaptively "smooth" zstat images (from GLM analysis or ICA equivalents;
the amount being controlled by beta priors) - with more smoothing in areas
where smoothness is warranted and less in areas where it is not (e.g. at
the interface between activated and null regions).
Cheers,
Andreas
Von: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of
"Anderson M. Winkler" <[log in to unmask]>
Antworten an: FSL - FMRIB's Software Library <[log in to unmask]>
Datum: Donnerstag, 8. Oktober 2015 09:38
An: <[log in to unmask]>
Betreff: Re: [FSL] Spatial smoothing for brainstem fMRI
Hi David, hi Andrew,
Very interesting paper. I'd say it's fine, no problems in having different
smoothing levels for different regions.
If you use FEAT, the smoothness will be computed from the residuals, which
will be similar to the one of the target (higher smoothing in this case).
All the best,
Anderson
PS: In my earlier email, "homologous reasons" should have been "homologous
regions" (!)
On 7 October 2015 at 15:46, David V. Smith <[log in to unmask]>
wrote:
Hi Andrew,
We used Anderson's suggested approach in a NeuroImage paper last year
(Murty et al.):
https://web.duke.edu/adcocklab/resources/pdf/restingstatevtasn.pdf
Time-series data from midbrain nuclei were unsmoothed (i.e., 0mm) and
cortical targets were smoothed at 6mm.
Hope this helps.
Cheers,
David
On Oct 7, 2015, at 10:21 AM, Andrew Song <[log in to unmask]> wrote:
Thank you for the detailed response Anderson,
That did come across my mind, but I have been cautious as I haven't seen
studies in which two different spatial smoothing kernels have been applied.
Do you think there are any potential problems with using this method? I
can't think of any, but I just wanted to make sure that the approach is
logically/theoretically sound.
Thanks again!
|