Hi Michael,
Generating mcflirt-style parameter outputs is relatively
easy. Just run avscale with the --allparams option and
grep out the required lines for Rotation Angles and
Translations.
As for getting a good estimate of through plane
effects from such a small ROI - I'm just not sure as
it will depend on how good the data is and how
big the rotations are. Can you upload the data for
me to have a look at?
All the best,
Mark
On 15 Oct 2008, at 13:00, Michael Hanke wrote:
> On Wed, Oct 15, 2008 at 10:10:57AM +0100, Mark Jenkinson wrote:
>> Hi Michael,
>>
>> I would try scripting flirt and see how that performs. In general
>> flirt will be more robust to things like limited FOV so hopefully
>> that
>> will help. Use the -nosearch option with this kind of data though,
>> as
>> that will help it.
> Thanks for your suggestions.
>
> I took the 4d image apart and ran flirt on each 3d image
> individually --
> it is not yet perfect, but when using -2D, a lot more stable than
> before. It gets even better when using the deweighting approach
> suggested by Andreas (using an mask covering the medial parts of the
> brain, excluding two edge slices on top and bottom).
>
> The thing it still cannot deal with are slight rotations around x-axis
> (ie. nods) -- quite plausible with -2D. However, doing -dof 6 it
> does not
> yield a stable alignment across the timeseries, regardless of
> deweighting enabled or not.
>
> Is there anything I could try to handle the nods?
>
> I am also interested in generating parameter output as mcflirt's -
> plots
> option would do -- is there a way to do that on the command line?
>
>
> Thanks again,
>
> Michael
>
>
>
>> On 15 Oct 2008, at 10:00, Michael Hanke wrote:
>>
>>> Hi,
>>>
>>> I have a dataset with 12 ZOOM-EPI slices (FOV 13cm), parallel to the
>>> calcarine sulcus, covering approx. 2/3 of the brain in
>>> ant.-post-direction.
>>>
>>> There is a slight drift in the data along the y axis, which I want
>>> to
>>> remove with mcflirt. However, mcflirt doesn't really like the data,
>>> ie.
>>> after correction, there are all kinds of additional (visible) motion
>>> in
>>> the dataset. I tried -2d, -gdt, -edge, -fov options, but could not
>>> achieve significant improvements. Especially the slow drift still
>>> remains in the data.
>>>
>>> I guess the problem is that the FOV barely covers the brain from
>>> left
>>> to
>>> right and only has the posterior parts. (screenshot is attached)
>>>
>>> Does anyone have experience with this kind of data and maybe
>>> suggestions
>>> what to try next? The amount of motion is more than I could tolerate
>>> for the intended analysis, so I cannot simply ignore it.
>>>
>>> Thanks in advance,
>>>
>>> Michael
>>>
>>> --
>>> GPG key: 1024D/3144BE0F Michael Hanke
>>> http://apsy.gse.uni-magdeburg.de/hanke
>>> ICQ: 48230050
>>> <zoom-epi.png>
>
> --
> GPG key: 1024D/3144BE0F Michael Hanke
> http://apsy.gse.uni-magdeburg.de/hanke
> ICQ: 48230050
>
|