Hi,
I've got a methodological question about the order of the procedures in fmri images procesing.
I was wondering why smoothing should be done after realigning and normalizing and not before. First, realignment of smoothed images should be more robust because high frequency noise is removed. No other effect should be produced if all the voxels are convolved with the same gaussian kernel. Second, images are normalized to a very smooth template. As I understand it is easier (in terms of time computing) normalizing smoothed image to a smoothed template than doing to non smoothed images; and the result shouldn't be affected.
Is there any theoretical reason for using this order of procedures? Is the order smoothing->realign->normalize wrong or can both of them be used indistinctly?
Thanks,
Carles Falcon
SCT-UB / CDI - Hospital Clínic Barcelona
Lab. Biofísica i Bioenginyeria. F. Medicina.
Universitat de Barcelona (SPAIN)
tf (34) 934 024 515
email: [log in to unmask]
|