Hi,
Be very careful to distinguish between a flip in A-P only vs a flip in
A-P *and* L-R.
The latter case still retains the same coordinate-system handedness (what we
call "radiological" or "neurological") and can be brought into correct
alignment
using FLIRT, but if it is *only* A-P that is flipped then the result of
registration
may look OK but actually by L-R flipped, as it cannot undo a single
flip, just
change which axis the flip is about.
I hope this helps.
All the best,
Mark
Vina Goghari wrote:
> Hi Mark,
> Thanks for the advise. The only thing that is flipped is
> Anterior-Posterior. However, it seems that if I allow for a full
> search for registration this does not cause a big problem.
>
> Thanks,
> Vina
>
> Mark Jenkinson wrote:
>> Hi,
>>
>> In Analyze you cannot store orientation information reliably (that's why
>> we prefer nifti). FSL has historically adopted the same conventions for
>> all Analyze images and these are the same as used in the avg152 images,
>> which can be described as "radiological" but that sometimes has
>> different
>> interpretations by different people. See our FAQ for the precise
>> definition.
>>
>> What I advise is to check the image by looking at it in FSLView, as this
>> will show you the labels such as L, R, A, P, S, I. If these are
>> correct then
>> everything is fine. If the acquisition is not axial then these will
>> be wrong
>> but check the example_func2standard image and see if that is correct.
>> If in doubt, run a test by doing something to the input image, inside
>> the brain,
>> such as manually editing some voxels to make them very dark which
>> you can then use to mark left and right and see if that is propagated
>> correctly.
>>
>> All the best,
>> Mark
>>
>>
>> Vina Goghari wrote:
>>> Hello,
>>> I'm working on a dataset collected a while back, where all the data
>>> was stored as Analyze, but none of the orientation information was
>>> stored in .hdr file. However, the data was stored in radiological
>>> convention (right first then left). I was just wondering how FSL
>>> handles this data with out info in the header and whether I have to
>>> worry about the data being flipped. In the end the FSL post-stat
>>> maps appear to also be in radiological convention.
>>>
>>> I hope this makes sense!
>>>
>>> Thanks,
>>> Vina
>>>
>
|