Dear Hauke,
much of the discussion on whether or not to perform slice-timing
correction is based on very fundamental philosophical considerations,
i.e. if you know that there is a well-defined change in the data (e.g.
each slice is shifted by 50ms) then it would be preferable to adjust
your model rather than interpolate your data. This would be the
preferable approach if you can ensure that your data is not mixed across
slices.
However, the various preprocessing procedures (i.e. motion correction,
normalization and smoothing in particular) introduce this mixing across
slices and thus model-based correction methods (with slice-dependant
models) might fail.
In the paper mentioned (Sladky et al, NI, 2011) we simulated various
paradigm designs and found that slice-timing correction works well over
a very wide range of parameters, for interleaved and sequential
acquisition. Therefore, I would recommend using slice-timing correction
unless you have a very good reason not to do so.
All the best,
Chris
--
Assoc. Prof. Priv.-Doz. Dr. Christian Windischberger
Vice Scientific Director
MR Center of Excellence
Center for Medical Physics and Biomedical Engineering
Medical University of Vienna
Lazarettgasse 14
A-1090 Vienna, AUSTRIA
Tel: +43-1-40400-6463
Fax: +43-1-40400-7631
Pieps: 81-6090
Am 27.07.2011 15:23, schrieb Hauke Hillebrandt:
> Dear SPM community,
>
> I have a question regarding slice timing correction in DCM/GLM.
>
> There was a recent paper suggesting that for GLM analysis it is better to include slice timing correction during preprocessing than to not include it:
>
> Neuroimage. 2011 Jul 2. [Epub ahead of print]
> Slice-timing effects and their correction in functional MRI.
> Sladky R, Friston KJ, Tröstl J, Cunnington R, Moser E, Windischberger C.
> http://www.ncbi.nlm.nih.gov/pubmed/21757015
>
> However, for DCM the recommendation is to specify the slice timings of each area and not use slice time correction during preprocessing:
>
> Neuroimage. 2007 Feb 15;34(4):1487-96.
> Dynamic causal modeling: a generative model of slice timing in fMRI.
> Kiebel SJ, Klöppel S, Weiskopf N, Friston KJ.
> http://www.ncbi.nlm.nih.gov/pubmed/17161624
>
> So, my question is whether it's better to use coordinates and extract VOIs from slice timing corrected or uncorrected data for DCM since the DCM analysis is dependent on the GLM?
>
> Best wishes,
>
> Hauke
>
|