Hi,
To add to this. FIRST gives you the segmentation in the native space of
the image. So, despite it has its own alignment procedure to MNI space,
this may be applied to the original image or the resampled/transformed
image.
The best way to transform the segmentations between spaces is to transform
the surfaces, in this way you are not susceptible to interpolation errors.
The surfaces may be warped through a non-linear warp field or transformed
by an affine matrix. Unfortunately, the software to warp/transform the
surfaces is not released.
Cheers,
Brian
> Hi,
>
> The easiest way to do that (and the only way if you are applying
> any non-linear transformations) is to take the resampled image
> (that is now in MNI space) and use the run_first script (as explained
> in the docs) but using an identity matrix for the transformation.
>
> Note that this isn't ideal, and if you are only doing affine
> transformations
> then you are better using the original image and a transformation
> matrix which describes the transformation to MNI space, but this
> is harder to create from other software outputs if you are not using
> flirt.
>
> All the best,
> Mark
>
>
>
> On 13 Dec 2008, at 20:46, siamak wrote:
>
>> Hi,
>>
>> I want to test my registration method and measure its effectiveness
>> in the
>> segmentation stage. I use an affine transformation to the MNI space
>> in my
>> first stage of registration as FIRST but my second stage of
>> registration is
>> different from FIRST. So, I have the registered image. Now, I only
>> want to
>> segment this image. FIRST tool combines registration and
>> segmentation. So,
>> my question is: How can I use segmentation part of FIRST alone?
>>
>> Thanks for your time,
>> Siamak
>>
>
>
|