Hi,
first, melodic does _not_ convert to %BOLD signal change unless you switch this on - are you sure you're running a recent version of FSL?
Wrt fsl_regfilt: this is designed to allow you to use the _time series output_ from melodic to denoise the original data. The further processing within melodic simply improves the characterisation of the temporal modes. These will then be regressed out of the _original_ data or any version thereof that you've prepared by any means you fancy. Please see the fslcourse and go through the course if you are unsure about how to use this.
hth
Christian
On 18 Dec 2012, at 17:57, Joseph Viviano <[log in to unmask]> wrote:
> I think you fail to understand my question... how am I supposed to do this properly? I could variance normalize all of my data pre melodic, but I would rather do that after averaging. But the prewhitening step eludes me.
>
> So, what should I do? Input registered, non normalized data into melodic, then independently convert to % signal change and variance normaize, and THEN use fsl_regfilt? Is there no way to apply melodic results to non-variance normalized data? What, exactly, should I do before I run melodic the first time, and before the second time? Especially since the --varnorm and --pbsc flags fail to actually turn off mean/variance normalization, so occasionally the program gets stuck trying to divide by 0...
>
> Please provide some sort of information on how to use your tool properly. There is so little documentation and I have no clear idea of what is actually happening to my data...
|