AFNI program: 3dCompareAffine
Output of -help
Usage: 3dCompareAffine [options] ~1
This program compares two (or more) affine spatial transformations
on a dataset, and outputs various measurements of how much these
transformations differ in spatial displacements.
One use for this program is to compare affine alignment matrices
from different methods for aligning 3D brain images.
Transformation matrices are specified in a few different ways:
* ASCII filename containing 12 numbers arranged in 3 lines:
u11 u12 u13 v1
u21 u22 u23 v2
u31 u32 u33 v3
* ASCII filename containing with 12 numbers in a single line:
u11 u12 u13 v1 u21 u22 u23 v2 u31 u32 u33 v3
This is the '.aff12.1D' format output by 3dAllineate,
and this is the only format that can contain more than
one matrix in one file.
* Directly on the command line:
'MATRIX(u11,u12,u13,v1,u21,u22,u23,v2,u31,u32,u33,v3)'
-------
Options
-------
-mask mmm = Read in dataset 'mmm' and use non-zero voxels
as the region over which to compare the two
affine transformations.
* You can specify the use of the MNI152 built-in template
mask by '-mask MNI152'.
* In the future, perhaps other built-in masks will be created?
*OR*
-dset ddd = Read in dataset 'mmm', compute an automask from
it (via program 3dAutomask), and use that mask
as the spatial region for comparison.
* If you don't give EITHER '-mask' or '-dset', then
this program will use an internal mask derived from
the MNI152 template (skull off).
-affine aaa = Input an affine transformation (file or 'MATRIX').
*OR* * You can give more than one '-affine' option to
-matrix aaa input multiple files.
* You can also put multiple filenames after the
'-affine' option, as in '-affine aaa.aff12.1D bbb.aff12.1D'
* The first matrix found in the first '-affine' option
is the base transformation to which all following
transformations will be compared.
------
Method
------
1) The input mask is hollowed out -- that is, all nonzero mask voxels that
do NOT neighbor a zero voxel are turned to zero. Thus, only the 'edge'
voxels are used in the computations below. For example, the default
MNI152 mask has 1818562 nonzero voxels before hollowing out, and
has 74668 after hollowing out. The hollowing out algorithm is described
in the help for program 3dAutomask.
2) For each surviving voxel, the xyz coordinates are calculated and then
transformed by the pair of matrices being compared. Then the Euclidean
distance between these two sets of transformed xyz vectors is calculated.
The outputs for each comparison are the maximum distance and the
root-mean-square (RMS) distance, over the set of hollowed out mask voxels.
The purpose of this program is to compare the results from 3dAllineate
and other registration programs, run under different conditions.
-- Author: RWCox - Mar 2020 at the Tulsa bootcamp
This page auto-generated on
Tue Dec 17 06:52:33 PM EST 2024