Brian,
I would think that Option One could be improved by *not* thresholding
the statistic images, but rather transforming the unthresholded image
and then applying a threshold to that transformed image. (Also be
careful about NaN's, as they 'bleed' when interpolated; make sure that
any image that you transform has zeros and not NaN's in the
background; to change NaN's to zeros see
http://www.sph.umich.edu/~nichols/JohnsGems.html#Gem2 )
In general, I would expect Option Two to provide slightly better
statistics because interpolation/resampling slightly smooths the data,
increasing SNR and reducing the multiple testing problem slightly.
Those improvements, of course, come at the expense of spatial
resolution, so it's not really an apples-to-apples comparison.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
On Mon, 1 May 2006, Lenoski, Brian - SJHMC wrote:
> Hi All,
>
> I seem to have two options when coregistering and reslicing
> functional and anatomical data sets. I need a thresholded spmT map
> in the same image space as the high-res anatomical upon completion
> of analysis.
>
> Option One:
>
> Coregister a saved, thresholded spmT map to the anatomical image
> using the mean epi image to obtain coreg/reslice parameters
>
> Option Two:
>
> Coregister and reslice entire realigned epi data set to anatomical
> image using mean epi image to obtain parameters. Then run stats on
> the smoothed resliced epi images resulting in spmT map in anatomical
> space.
>
> Visually, option two seems to produce sharper results but also takes
> exponentially longer to complete the analysis due to the increased
> complexity. My stats are improving but still pretty weak at this
> point so I was wondering if anyone has a mathematical reason/opinion
> for choosing one method over the other assuming time is not a
> factor.
>
> Thank you,
>
> Brian
>
>
>
>
>
>
>
>
>
>
>
> Brian Lenoski
>
> Neuropsychology Neuroimaging Laboratory
>
> Barrow Neurological Institute
>
> 222 West Thomas Rd, Suite 310
>
> Phoenix, AZ 85013
>
>
>
> ph 602.406.4494
>
> fax 602.406.3498
>
> email [log in to unmask]
>
>
>
>
|