[MINC-users] Automated Segmentation Yet More Questions

[Paul GRAVEL]

Hello all,

Thank you very much for your feedback.
It does however raise a few questions.
(Please see below for previous replies to previous questions).

1 - Since ANIMAL uses the ICBM152 model for more robust linear
    registrations, does this mean that "mritotal" uses
    the ICBM152 model by default as well, or should I use the
    "-model icbm_avg_152_t1_tal_lin"?

    e.g. mritotal -model icbm_avg_152_t1_tal_lin $arg1'_mri.mnc' $arg1'_mri_stx_lin.xfm'

2 - If the "-model" switch needs to be used, is the
    "icbm_avg_152_t1_tal_lin" file the correct option to use
    for the ICBM152 model?

3 - After segmentation, in order to remove voxels in the skull,
    should I use a different mask than "average_305_mask.mnc"?

Once again, I thank you.

Best Regards,

Paul

On Mon, 21 Jun 2004, D. Louis Collins wrote:

> Paul,
>
> The procedure described on my web site is still valid, however, many of
> the programs have been updated and yield better results now.
>
> Answers are below.
>
> -Louis
>
>
> On Jun 21, 2004, at 11:07 AM, Paul GRAVEL wrote:
>
> > Hi All,
> >
> > I am not sure if this is the appropriate mailing list to send these
> > questions(bottom of e-mail) to, but I will ask them anyway.  Please
> > let me
> > know if I should send them to another mailing list.
> >
> > I have 42 subjects that I needed to segment into several ROI's.
> > I used the procedure displayed on Dr. Louis Collins Web Page
> > (http://www.bic.mni.mcgill.ca/users/louis/pet_segment)
> >
> > The procedure is as followed:
> >
> > Step 1: mritotal $arg1'_mri.mnc' $arg1'_mri_stx_lin.xfm'
> >
> > Step 2: nu_correct $arg1'_mri.mnc' $arg1'_mri_nuc.mnc'
> >
> > Step 3: mritotal -nonlinear $arg1'_mri_nuc.mnc' -transformation
> > $arg1'_mri_stx_lin.xfm' $arg1'_mri_stx_nonlin.xfm'
> >
> > Step 4a: mincresample -like
> > /avgbrain/brain/images/icbm_template_1.00mm.mnc
> > $arg1'_mri_nuc.mnc' -transformation $arg1'_mri_stx_lin.xfm'
> > $arg1'_mri_stx_nuc.mnc'
> >
> > Step 4b: classify_clean $arg1'_mri_stx_nuc.mnc'
> > $arg1'_mri_stx_classes.mnc'
> >
> > Step 5: stx_segment $arg1'_mri_stx_nonlin.xfm' $arg1'_mri_stx_lin.xfm'
> > $arg1'_mri_stx_classes.mnc' $arg1'_mri_stx_labels.mnc'
> >
> >
> > My questions are as followed:
> >
> > 1 - Is this still the best procedure to do segmentation?
>
> pretty much.  ANIMAL now uses the ICBM152 model for more robust linear
> registrations.
>
> > 2 - Is step 2(nu_correct) based on the N3 Algorithm, if not which
> > algorithm is it based on?
>
> this is based on N3, the method of John Sled.
>
> > 3 - Is step 4b(classify_clean) based on the INSECT (Intensity
> > Normalized
> > Stereotaxic Environment for the Classification of Tissue) Algorithm, if
> > not which algorithm is it based on?
>
> classify_clean is a perl script that uses both bayesian and artificial
> neural nets for classification and is based on INSECT.
>
> > 4 - Is step 5(stx_segment) based on the ANIMAL (Automatic Nonlinear
> > Imaging Matching and Anatomical Labelling) Algorithm, if not which
> > algorithm is it based on?
>
> step 5, stx_segment, takes the output of the linear and non-linear
> registrations and combines it with the classified data to achieve a
> segmentation.  The entire procedure, registrations+classifications is
> ANIMAL.
>
> Hope this helps,
>
> -L
>
>
> >
> > I thank you very much in advance.
> >
> > Best Regards,
> >
> > Paul
> > ______________________________
> >
> > Paul Gravel
> > Neurobiological Psychiatry Unit
> > McGill University
> > 1033 Pine Avenue West
> > Montreal, Quebec, Canada, H3A 1A1
> > Phone:  (514) 398-7301
> > Fax:    (514) 398-4866
> >
> >
> >
> >
> > _______________________________________________
> > MINC-users@bic.mni.mcgill.ca
> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>

______________________________

Paul Gravel
Neurobiological Psychiatry Unit
McGill University
1033 Pine Avenue West
Montreal, Quebec, Canada, H3A 1A1
Phone:  (514) 398-7301
Fax:    (514) 398-4866