Hi,
Sorry for the delay - I've just come back from an
email-free Easter holiday.
In general I'm not sure that this will work. Large/fast
motion is often accompanied by a range of artefacts
(e.g. blurring, spin history, through-slice effects, etc)
so that it is difficult/impossible to get good registrations.
If you do want to persist then it would be possible to
add the 16mm and 0.5mm scales, but for 0.9mm
resolution I see no point to the 0.5mm scale as 1mm
will be fine. If it is currently getting "close" and you
want to see if you can squeeze more accuracy from
it then I would do a second stage registration with
only the 1mm scale with the -init, -nosearch, -noresample
options. This will let it use the native 0.9mm image.
Adding the 16mm stage might help if the current
method is not getting close. However, you might
find that it does not help if your FOV is not large.
I have found in the past that things like brain
extraction are more important than adding other
sampling stages.
If you are convinced that having the 16mm might
help then I can show you how to add it - just let me
know.
All the best,
Mark
On 4 Apr 2010, at 18:58, Junqian Gordon Xu wrote:
> Mark,
>
> I have to accommodate both large erratic motions and a 0.9x0.9mm in-
> plane resolution.
>
> Could you teach me how to modify sch2D_3dof to let it initiate the
> search from 16 mm scale and optimize down to 0.5 mm scale (i.e. 16mm-
> >8mm->4mm->2mm->1mm->0.5mm), if the latter makes sense to you. I'm
> pretty sure I need the former after some experiment with flirt -2D.
>
> Thanks
> Gordon
>
> On 03/31/2010 03:16 AM, Mark Jenkinson wrote:
>> Yes, although not easily with the default set of schedule file.
>> It requires the existing xyztrans.sch to be modified to remove
>> the z translation.
>>
>> It is very easy, so I've done that and am including the
>> text below. Save this, as plain text, in a file and then use
>> the -schedule option with this file and you can get your xy
>> translation.
>>
>> All the best,
>> Mark
>
|