Hi Jack,
This isn't easy to answer in general because it does depend on how good the
images look and how good/bad the current registration is given the level
of artefacts etc.
What we recommend is that to get good registration from the subject's
individual to standard space, the "best" image (high res & good contrast)
is used last in the chain (i.e. the Main Structural).
For the T2* (EPI?) to structural, normally the same structural is fine
without
an Initial Structural. The latter is usually only needed when there is
a partial
FOV in the functional images (in which case you put a whole brain EPI in as
the Initial Structural). You can of course try using your SPGR or T1
images in the orders you suggest and one might turn out better, but normally
there isn't much to be gained. The most significant thing in EPI to
structural
registration is usually the susceptibility distortions which can be
alleviated
with FUGUE *if* you have acquired field maps.
Sorry I can't be more definitive.
Cheers,
Mark
Jack Grinband wrote:
>Thanks, I'll go back to the simpler approach.
>
>I have one more question. I'm trying to get a better registration of my functional data.
>Would it help if I added another registration step? i.e. switch from
>T2* --> T1, low res --> MNI152
>to
>T2* --> T1, low res --> SPGR, high res --> MNI152
>or even
>T2* --> SPGR, low res --> SPGR, high res --> MNI152
>
>Since the SPGR images have a much higher S/N and contrast, one would expect
>better registration. On the other hand, the T2* has such low contrast that maybe it
>won't help at all. Since I'm not really sure how good the registration should look, it's
>hard to know when to stop tweaking the data.
>cheers,
>
>jack
>
>
>
>On Sat, 26 Apr 2003 16:06:37 +0100, Mark Jenkinson <[log in to unmask]>
>wrote:
>
>
>
>>Hi Jack,
>>
>>We tested this idea when we first wrote mcflirt.
>>Somewhat counter-intuitively it did worse (though not hugely) than
>>using the single volume. So we went for the simple, and better, approach.
>>
>>One reason that it might not help is that some artefacts in the images may
>>end up blurring the images over a greater area than you'd get from just any
>>one instance.
>>
>>As for your commands, they look just fine.
>>Is there a problem with the motion correction results?
>>If so, then the average won't be too nice.
>>
>>All the best,
>> Mark
>>
>>
|