[MINC-users] best_nonlin_reg.pl ?

[Andrew Janke]

On Sat, Apr 9, 2011 at 09:28, Lisa F. Akiyama <lrisa87 at uw.edu> wrote:
> Another question on making average models....
>
> I've been creating several preliminary models using different configurations
> and methods.
> The MNI average brain models for both adults and pediatrics are all straight
> in every direction.
> However, I look at my resulting models, some of them are a tad tilted in one
> or more directions. Going back and checking closely the individuals' scans
> that were coregistered (linear 9 parameter registration) to a standard
> template,

This depends on whether you making a symmetric or non-symmetric
(default) model. If non-symmetric the resulting image will be in the
average space of all your input images meaning there is a left/Right
rotation bias when your subjects are being scanned.

If you make a symmetric model (add -symmetric to volgenmodel) then
this should not happen.

> Is this a common issue faced when coregistering each individual to a
> standard template and creating a model?

Yes, you are not alone!

> If so, what are some of the common
> ways of fixing them so that the resulting average model will be straight in
> every direction and still keeping
> the proper MNI space?

I have a quick and dirty tool called volalign that does a registration
of an image to itself and then takes 1/2 the resulting translation and
rotation results and applies it. It won't be perfect as you really
should iterate this sort of approach but with one run it will make
things a whole lot better.

You can get it here:

   http://packages.bic.mni.mcgill.ca/scripts/volalign

Use it like this:

   $ volalign rotated-model.mnc align.xfm aligned-model.mnc

Both align.xfm and aligned-model.mnc are output files.

As part of this you might also then like to align 0,0,0 about the
centre of the volume (or sometimes before), to do this use volcentre:

   http://packages.bic.mni.mcgill.ca/scripts/volcentre

(sorry Nth Americans, you're going to have to learn how to spell, the
fact that we English stole the word from the French is irrelevant)

Use it as such:

   $ volcentre -zero_dircos in.mnc out.mnc

use -help for more options.

Then when you use volalign you can use the -world option so that your
resulting image is centered about the 0,0,0 world co-ordinate.


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