Hi,
first_flirt is the processing that was a applied to all images prior to
constructing the shape model.
The shape variation modelled is residual to this alignment process. This
is why it is suggested
that first_flirt is used for the alignment. What the program "first" cares
about is a transformation matrix, this is used to
transform the model into the original space (hence no re-sampling of the
image). The model registration is called within
the FIRST code (first.cc) and transformation is is implemented in the
"shapeModel" class.
Cheers,
Brian
> Hi Mark,
>
> I did some pos tprocessing and I think I have problem again in this step.
> So I
> came back to this question. I checked the scripts of run_first_all and
> first_flirt. It seems that the matrix is the concatened matrix of
> registration in
> stage 1 and stage2. So, you mean that I should use only mat_stage1.mat or
> only mat_stage2.mat or -concat <mat_stage1.mat>, <mat_stag2.mat>?
>
> I have another question that may be related to above again. FSL does the
> segmentation on images that their dimention and orientation is not
> necessary
> the same as MNI152_T1_1mm. The dimention and orientation of the first and
> second steps of registration change to MNI reference image as I checked in
> first_flirt (by debugging), but the final segmented image has the same
> orientation and dimention of the original image. I tracked the scripts
> run_first_all, first_flirt and run_first. I didn't find any step that this
> resampling
> to original grid and orientation occurs. Could you please guide me in this
> regard?
>
> Siamak
>
>
|