[MINC-users] labeling
Daniel Schwarz
schwarzd@feec.vutbr.cz
Wed Nov 24 02:25:04 2004
Thank you very much, Louis.
Minimizing the errors = my next step. I would like to try your
stx_segment, I am waiting patiently for the time, when it is available
for public. I am also trying some other methods to find deformation
field which may yield better fit. Could you please explain briefly the
role of the "heuristic rules" in stx_segment?
Thank you.
Daniel
D. Louis Collins napsal(a):
>Daniel,
>
>You have it exactly - this step customizes a standard set of labels
>(phantom_labels.mnc) to your subject.
>
>The script stx_segment uses this as a basic step, but combines the result
>with classsification results and heuristic rules to minimize errors.
>
>-Louis
>
>
>On 11/23/04 10:49 AM, "Daniel Schwarz" <schwarzd@feec.vutbr.cz> wrote:
>
>
>
>>Dear minc-users,
>>
>>I would like to ask one thing, just tu make sure I don't do nonsense.
>>Is this labeling?
>>
>>mincresample -nearest_neighbour -invert_transform -transform
>>nonlinear.xfm -like image.mnc phantom_labels.mnc labeled_image.mnc
>>
>>where:
>>
>>- image.mnc is the image I want to segment/label.
>>
>>- nonlinear.xfm is file containing the affine transform and the name of
>>the file with displacement field. I have got it with MRITOTAL and
>>MRITOTAL -nonlinear, where the model was phantom_template.
>>
>>- phantom_labels.mnc is segmented phantom_template.mnc
>>
>>- labeled_image.mnc is the expected segmented/labeled image.mnc
>>
>>
>>Thanks a lot for your answers/corrections/notices.
>>
>>Daniel Schwarz
>>
>>
>>
>>_______________________________________________
>>MINC-users@bic.mni.mcgill.ca
>>http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>>
>>
>
>
>