[MINC-users] best_nonlin_reg.pl ?

[Andrew Janke]

Hi again Lisa,

> I am trying to create average infant head templates similar to what Fonov et
> al 2011 have created with the pediatric population.
> (He informed me that the infant templates available online were created
> using similar methods as stated in his 2011 paper involving pediatric data)

> Thanks for guiding me on how to modify the configurations.
> As a start, I will play around with nlpfit and fit config (I'll use what
> Fonov et al used for their settings as a guidance).

OK, now I am with you. Vladimir (Fonov) wrote a ruby? script that
behaves much as 'volgenmodel'
(http://packages.bic.mni.mcgill.ca/scripts/volgenmodel)  does. The
difference as he mentions in his paper is that he doesn't start
fitting from scratch at each step. volgenmodel does but there are
other reasons for this.

You are welcome to use it as a start if only because last I looked at
Vladimir's scripts they use a lot of his own mashups of others code
(mincANTS, ITK etc). In time you should be able to get his version
working but I suspect volgenmodel will make a good starting step.

It has a few dependencies on other scripts that I have also uploaded
to the scripts directory, note that a run of this script will take a
LONG time if you dont run it on a batch (SGE/gridengine) setup.  For
example a run on 200 elderly brains takes approximately 1 week on a
50core linux cluster.

Usage is like this:

   $ volgenmodel -check -clean -symmetric -normalise -extend 5 \
      -output_model <model.mnc> -workdir genmodel-work \
      <infiles>

The -check option generated JPG files in the workdir, -clean removes a
bunch of working files as it goes along, -symmetric makes a symmetric
model, -normalise takes care of global intensity differences between
images, -extend 5 adds a 5 voxel boundary around all the images in
case the cropping/framing is off.

Note that for 100 subjects this will need about 700GB of spare space
in the -workdir directory if the -clean option isn't used.

Good luck.


-- 
Andrew Janke
(a.janke at gmail.com || http://a.janke.googlepages.com/)
Brisbane->Australia    +61 (402) 700 883