[MINC-users] Atlas coregistration using minctools

[Andrew Janke]

Hi Aaryani,

On Thu, Jun 9, 2011 at 03:59, Aaryani Tipirneni <atipirne at stanford.edu> wrote:
>  I'm working on affine coregistering of the 3D data to the atlas. Sometimes using just mritotal works fine but sometimes it is failing. Can anyone suggest more robust ways of coregistering using minc tools.

mritotal is specifically designed to register T1 human data to the
ICBM/MNI averages. Claude Lepage and myself have released a few
scripts that take a more general approach to linear and nonlinear
registration. You can find these two scripts here:

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

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

Both of these don't require a directory of model files and are made to
work either between individual images or to a model.  ie:

   $ bestlinreg source.mnc target.mnc out.xfm [resampled.mnc]

I only wish that Claude chose a better name for bestlinreg as I can't
imagine what he's going to call the next version!

>  I have one more question regarding affine registration; I performed affine registration to two data sets (at two different time points) of same subject to the atlas and they seemed to coregister well. But if I remove the scaling (setting scaling in xfm2param to 1) to the 2 data sets to avoid zoom -in/out, do you still expect them to coregister well? I see subtle differences in the coregistration as each data set is coregistered independently to the atlas and depends on their transformation matrices (combination of translation, rotation, scaling, shear).

I don't quite follow what you mean by "setting scaling in xfm2param to
1", however what I think you want to do is change the input parameters
to minctracc.

   -lsq6 = rotation and translation only
   -lsq7 = lsq6 + one global scaling
   -lsq9 = lsq6 + 3 independent scales

You can also add these as an argument to bestlinreg.

Good luck!


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