Hmm... actually, this didn't achieve any actual results. So, there's
still a kink somewhere. Any ideas?
Thanks,
Dav
On Nov 8, 2009, at 8:28 PM, Dav Clark wrote:
> For posterity (and correcting me if I'm wrong!), it seems that all
> one needs to do is add a reg directory in each lower-level feat
> directory, ie:
>
> scan1.feat/reg
> scan2.feat/reg
> etc.
>
> Then, all you need there is an example_func2standard.mat, and a
> standard.nii.gz (a copy of the standard image you'd like to use,
> like MNI152_2mm_brain). Perhaps there are equivalent things you
> could do for a highres or similar. For completeness, the .mat file
> should be:
>
> 1 0 0 0
> 0 1 0 0
> 0 0 1 0
> 0 0 0 1
>
> Then, if you do a higher level model on these feat directories,
> those files will be used to coregister your copes, etc. I suppose
> you may need to delete reg_standard directories if they were there,
> but they seem to be generated by the higher-level analysis
> automatically if they are not there (at least for FSL 4.1.4).
>
> The only thing I'm a little concerned about is that trilinear
> interpolation is still going on here. I'm not sure what trilinear
> interpolation does for an identity matrix... but I imagine I might
> still be getting a bit of extra smoothing.
>
> Cheers!
> Dav
>
> On Nov 6, 2009, at 12:06 AM, Stephen Smith wrote:
>
>> Hi, yes you can overwrite the relevant .mat file(s) in the reg
>> directory, delete the reg_standard directory (if it exists already)
>> before running featregapply again.
>> Cheers.
|