I completely agree! But then again some scanners suffer from serious N/
2 ghosting and then you would actually like to get the aliased eye
movements removed from the visual cortex.
Best
Torben
Den 02/04/2009 kl. 10.41 skrev Michael T Rubens:
> Perhaps knowing a bit about the type of task would help determine
> the best method. For many visual tasks it is likely that the signal
> you would extract from the eyes would highly correlate with the
> task, thereby regressing out signal of interest and killing your
> power.
>
> -Michael
>
> On Thu, Apr 2, 2009 at 1:23 AM, Torben Ellegaard Lund <[log in to unmask]
> > wrote:
> Hi Dorain
>
> There are several ways this could be done, and which one you choose
> should depend on your programming skills. But in general you should
> have a pretty good prior hypothesis that the particular area you
> want to mask out by regression will be a noise only area. But this
> would e.g. apply to ventricles, major blood vessels and eyes.
>
> One way to do this is to use the Eigenvariate button, when you view
> your results. Move the cursor to the eyes, and use the Eigenvariate
> button to extract the timeseries from the region, or single pixel.
> Now reanalyse your data with the extracted timeseries entered a a
> covariate, much like you would do with motion parameters. The
> timeseries you want to remove will be found in the variable xY.u
>
> Alternatively you could find typical MNI space locations for eyes
> and ventricles and automatically extract timeseries from those
> regions using spm_sample_vol.m This would require a bit of matlab
> coding.
>
> The benefit of regression as opposed to masking is that noise could
> be removed from other areas than the ones where you timeseries was
> extracted from. The drawback is that you risk removing real signal,
> if the noise looks like the signal.
>
>
>
> Best
> Torben
>
>
>
>
>
>
>
> Den 02/04/2009 kl. 06.49 skrev Dorian P.:
>
> Dear Torben,
>
> How can a time series from a specific voxel be added as a regressor?
>
> And can this be done to covary out the activity similar to any voxel
> of artifactual activity?
>
> Thank you.
> Dorian.
>
> 2009/4/1 Torben Ellegaard Lund <[log in to unmask]>:
>
> Dear Alexander
>
> If the eye-artefact is only there in some of the con images it will
> not make
> it through the threshold in the final second level analysis. This
> could have
> been the case in a fixed effects analysis but not in a random effects
> analysis. If you want to avoid those artefacts you could include a
> time-series from an eye-voxel in your design matrix. This would most
> likely
> remove the eye-artefact, but you risk removing some of the activity
> as well.
>
> Best
> Torben
>
>
>
> Den 31/03/2009 kl. 16.56 skrev Alexander Lebedev:
>
> Dear SPM experts
>
> I decided to check my old results, and found one problem. When I have
> opened con*-files in xjview tbx, strange thing appears... There are
> activations of eye movements (notwithstanding of Normalization) in
> some
> con*-files. May I include such results in group study? Could you
> advice me
> any solutions to prevent this trouble?
>
> Thank you beforehand
>
> Best Regards
> --
> Alexander Lebedev.
>
>
>
>
> --
> Research Associate
> Gazzaley Lab
> Department of Neurology
> University of California, San Francisco
|