[MINC-users] classify_clean

[Mishkin Derakhshan]

Hi Paul,
Below is how I referenced the minc tools and classify_clean. Perhaps
we should have a page with a list of references and how to reference
the tools on the bic website (or the Minc wikibook) much like the
folks at FSL and freesurfer.
hth,
mishkin

Medical Imaging NetCDF (MINC) is a medical imaging data format
and associated set of tools and libraries developed at the Montreal
Neurological Institute (MNI) and freely available online (http://www.
bic.mni.mcgill.ca/ServicesSoftware). The tool classify_clean, which is
used to classify stereotaxic MINC volumes, involves a Bayesian labeling
scheme and a set of standard sample points to compute an initial volume
classification. This classification is then employed to purge incorrect tag
points from the standard set, yielding a custom set of labels for the
particular subject. Finally, this tag point set is used by an artificial neural
net classifier to classify the volume (Zijdenbos et al., 1998).

Zijdenbos, A., Forghani, R., Evans, A., 1998. Automatic quantification
of MS lesions in 3D
MRI brain data sets: validation of INSECT. Medical Image Computing and
Computer-Assisted Interventation—MICCAI'98, pp. 439–448

On Sun, Nov 6, 2011 at 7:33 PM, Paul Rasser
<Paul.Rasser at newcastle.edu.au> wrote:
> Thanks Andrew.
>
> re: -tagfile -bgtagfile, in a way yes, as I have tested with -tagdir to point to the custom inverted tags (which I call ntags_1000_bg.tag and ntags_1000_prob_90_nobg.tag) , but  -tag_transform looks like it might be a better option.
>
> Also, any help with the  reference for classify_clean and nlfit_smr would be greatly appreciated.
>
> Thanks again,
> Paul
>
>
>>>> Andrew Janke <a.janke at gmail.com> 04/11/11 4:51 PM >>>
> Hi Paul,
>
> On 4 November 2011 14:03, Paul Rasser <Paul.Rasser at newcastle.edu.au> wrote:
>> I'm looking to use classify_clean to classify volumes in native space and was hoping for suggestions/concerns with respect to using it in the following way:
>
> I generally always do tissue classification in model space but I can't
> see any reason as to why this approach won't work.
>
>> #find nl transformation
>> nlfit_smr subj.mnc nl.xfm -model colin27_t1_tal_lin
>>
>> #invert transformation
>> xfminvert nl.xfm  nl_inv.xfm
>>
>> #invert .tag files to subject's native space
>> transform_tags ~/ntags_1000_prob_90_nobg.tag nl_inv.xfm ntags_1000_prob_90_nobg.tag
>> transform_tags ~/ntags_1000_bg.tag nl_inv.xfm ntags_1000_bg.tag
>>
>> #apply classfy_clean to subject in native space
>> classify_clean -clean_tags subj.mnc subj_seg.mnc
>
> I take it you also supply -tagfile -bgtagfile arguments so that your
> custom tags are used?  Either that or you can always skip transforming
> the tags and instead supply a  -tag_transform <nl_transform.xfm>
> argument.
>
>
>
> a
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>