On Fri, 25 Jul 2008 22:17:49 +0100, <Austin> <Woolard>
<[log in to unmask]> wrote:
>Dear all,
>
>One of our subjects has substantial movement artifacts (probably due to a
>sneeze or cough during the MRI) in a couple of volumes. Is it possible to
>remove these specific volumes from our fMRI analysis? If this is possible,
how
>might I remove them from the analysis?
I basically agree with Helmut.
1. It's best to discard the data. Sometimes, based on a cost-benefit analysis
(especially in the case of subjects who are _not_ normal volunteers) that is
considered too "expensive".
2. In the model, "regress out" the bad volumes. Warning: you need a
separate regressor for each bad volume.
In addition, you might want to consider removing the volume _prior_ to
preprocessing, in the case you do slice timing correction. Why? Because slice
timing correction involves interpolation, and the problematic data might be
smeared through time into nearby volumes. Of course, you CANNOT simply
delete the volume, because there would be a "hole" in the time series. Rather,
you could replace it with something more benign, e.g. the average of its
neighbors.
>
>Thanks, in advance, for your help.
>
>Austin Woolard
>Psychiatry
>Vanderbilt University
|