In principle, the data only needs to be resampled once. Each time it is
resampled, it becomes slightly degraded. In practice though, it
probably makes relatively little difference for spatial normalisation
(unless it subsumes the motion correction).
The more important point is to ensure that spatial transformations that
include those for head motion correction, use a more accurate
interpolation method. B-spline interpolation tends to be more accurate
than trilinear interpolation (for images that can be considered Gaussian
processes), so if motion correction (maybe with unwarping) is done with
B-spline interpolation, the results should contain slightly less
motion-related artefact.
See the following paper for more information about effects of head
motion correction:
Comparison of fMRI motion correction software tools
NeuroImage, Volume 28, Issue 3, 15 November 2005, Pages 529-543
T.R. Oakes, T. Johnstone, K.S. Ores Walsh, L.L. Greischar, A.L.
Alexander, A.S. Fox, R.J. Davidson
But also note the following, which is a result of treating motion correction
and the GLM as two separate and sequentially applied tools - as opposed to
trying to model the data in a more coherent way.
Motion Correction Algorithms May Create Spurious Brain Activations in
the Absence of Subject Motion
NeuroImage, Volume 14, Issue 3, September 2001, Pages 709-722
L. Freire, J. -F. Mangin
For examples of the effects of repeated interpolation, see some of
Thevenaz's work.
Best regards,
-John
On Wed, 2010-02-24 at 07:44 +0000, Stephan Walther wrote:
> Hi,
>
> I decided to change my proprocessing with using unwarping and realised now
> that in both the unwarping and the normalisation fourth degree b-spline
> interpolation is default.
> I remember that at a SPM course we were told to use b-spline interpolation
> only one time and when the EPI images are resliced directly in the realignment
> step.
> Is this correct? What disadvantages are combined with using the b-spline
> interpolation two times benath a greater interpolation error? What are the
> alternatives for my preprocessing?
>
> thanks in advance,
> Stephan
>
--
John Ashburner <[log in to unmask]>
|