My 2¢ is that higher dimensionality is better for fsl_regfilt. But, as you
point out, reviewing too many components can be tedious and, at some point,
prone to error.
On Wednesday, October 19, 2011 10:29:51 you wrote:
> Hi FSL Experts,
>
> I have a quick question about using MELODIC to identify and remove (via
> fsl_regfilt) motion and other artifact components from functional EPI
> data:
>
> Is it better to leave on "automatic dimensionality estimation" and let
> MELODIC determine the number (~105-115) or turn it off (and choose 40
> components)?
>
> Doing it the latter would certainly be faster (as there would be about a
> 1/3 as many to components to review), but I am worried that it might
> increase the risk that artifacts would be grouped with task-related
> signal. Thanks!
>
> Cheers,
> Pete
>
> Dept. of Anatomy & Neurobiology
> Boston University School of Medicine
|