Just to add a bit of clarification -
Negative voxel sizes are a pain and I've tried to describe them and
how they affect things in the following pages:
http://www.fmrib.ox.ac.uk/fslfaq/#general_lr
http://www.fmrib.ox.ac.uk/fslfaq/LRfaq.html
It seems that Fast doesn't correctly preserve the sign of dimensions
and origins, but as Steve said, that's easy to sort out with avwcpgeom.
I do fully agree with Clare though - it is much better to run Fast on
the original image and then resample the outputs later.
Finally, Flirt doesn't actually care if there is a negative dimension
or not - it is effectively ignored - so if you are using it to indicate
left/right swaps in the way the data is stored then be aware that
Flirt won't be consistent with this. We only have the negative
voxel dimension for consistency with the standard image and use
it to detect this and calculate "Talairach" coordinates appropriately.
Flirt will also *never* swap left and right in an image during
registration (all registration matrices always have positive determinant).
If you need to do this the utility avwswapdim can be used - e.g.
avwswapdim imagein -x y z imageout
(depending on how your image is stored).
Hope this helps clarify all the issues a bit.
All the best,
Mark
Stephen Smith wrote:
>I agree with Clare here - must be a bug in FAST that it's not carrying
>over the -ve voxel dimension (which is part of our old standard way of
>defining Tal space). We'll fix that - in the meantime you can use
>avwcpgeom to reset the dims easily - but also I agree with Clare that it's
>much better to segment the original image not a transformed one.
>
>Cheers.
>
>
>On Thu, 22 Jul 2004, Clare Mackay wrote:
>
>
>
>>Hi Krish,
>>
>>Sounds like FAST doesn't keep the negative voxel dimension, but one
>>thing to consider is that you would probably be better off running FAST
>>on your data in their original space then registering the data to the
>>template afterwards. The interpolation that happens as a consequence of
>>registration won't help with the segmentation, and it's easy to apply
>>FLIRT transforms to the segmentations afterwards.
>>
>>Cheers,
>>Clare
>>
>>___________________________________________________
>>Clare E. Mackay, Ph.D.
>>Research Fellow,
>>Prince of Wales International Research Centre (POWIC),
>>University Department of Psychiatry,
>>Warneford Hospital, Oxford. OX3 7JX. UK
>>
>>tel: 01865 455910 fax: 01865 455922
>>OCMR: 01865 221866 / 72
>>[log in to unmask] www.psychiatry.ox.ac.uk/powic
>>___________________________________________________
>>
>>
>>>>>[log in to unmask] 07/22/04 3:43 PM >>>
>>>>>
>>>>>
>>Sorry everyone! This should read:
>>
>>"it seems as if the Talairach origin is not carried over by **Fast**
>>either".
>>
>>It is Fast that I am having problems with, NOT Flirt.
>>
>>Krish
>>
>>
>>On 22/7/04 3:17 pm, "Dr Krish Singh" <[log in to unmask]> wrote:
>>
>>
>>
>>>Just to add to my last post, it seems as if the Talairach origin is
>>>
>>>
>>not
>>
>>
>>>carried over by Flirt either.
>>>
>>>Krish
>>>
>>>
>>>
>>>
>>>>Hi there,
>>>>
>>>>I've got myself into some laterality difficulties with negative voxel
>>>>dimensions. Let me explain what I did:
>>>>
>>>>1) Using flirt I registered a radiological axial volume to the FSL T1
>>>>template.
>>>>avwhd now reports that this has voxel dimensions of -2,2,2 mm which
>>>>
>>>>
>>is
>>
>>
>>>>correct
>>>>I think.
>>>>
>>>>2) I took the registered volume generated in the above step and
>>>>
>>>>
>>passed it
>>
>>
>>>>through Flirt to generate segmentation maps. avwhd now reports that
>>>>
>>>>
>>the voxel
>>
>>
>>>>dimensions for the segmented images are 2,2,2 which I don't think is
>>>>
>>>>
>>correct
>>
>>
>>>>as
>>>>they should be the same as the input image.
>>>>
>>>>Is the above behaviour expected? It causes me some problems because
>>>>
>>>>
>>my
>>
>>
>>>>software, mri3dX, assumes that if you overlay one volume on another
>>>>
>>>>
>>and their
>>
>>
>>>>horizontal voxel dimensions have different signs, then it needs to
>>>>
>>>>
>>flip the
>>
>>
>>>>laterality of the overlay. In the case of the above, this is clearly
>>>>
>>>>
>>not the
>>
>>
>>>>case.
>>>>
>>>>All the best,
>>>>
>>>>Krish
>>>>
>>>>
>>>--
>>>Dr K.D. Singh
>>>Senior Lecturer and Convenor of the Neuroimaging Research Group,
>>>Director, MRI Research Centre,
>>>Neurosciences Research Institute
>>>Aston University, Birmingham B4 7ET, U.K.
>>>Tel/Fax: +44 (0)121 [204 3865/3864]/[333 4220]
>>>[log in to unmask], http://www.aston.ac.uk/lhs/staff/singhkd/
>>>
>>>
>>>
>>--
>>Dr K.D. Singh
>>Senior Lecturer and Convenor of the Neuroimaging Research Group,
>>Director, MRI Research Centre,
>>Neurosciences Research Institute
>>Aston University, Birmingham B4 7ET, U.K.
>>Tel/Fax: +44 (0)121 [204 3865/3864]/[333 4220]
>>[log in to unmask], http://www.aston.ac.uk/lhs/staff/singhkd/
>>
>>
>>
>
> Stephen M. Smith DPhil
> Associate Director, FMRIB and Analysis Research Coordinator
>
> Oxford University Centre for Functional MRI of the Brain
> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
>
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>
>
|