OK - very sorry, it did work. No more e-mails from me for at least
several hours now!
DC
On Nov 8, 2009, at 8:32 PM, Dav Clark wrote:
> 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.
|