bestlinreg_g

perform best linear registration (generalized variant)

bestlinreg_g [options] source.mnc target.mnc output.xfm [output.mnc]

DESCRIPTION

bestlinreg_g is an enhanced variant of bestlinreg.pl that adds support for selectable cost functions and additional transformation constraints. Like bestlinreg.pl, it performs hierarchical multi-scale linear registration between a source and target MINC volume using minctracc, proceeding through several blur and step-size stages.

This variant allows the user to choose among mutual information, normalised mutual information, and cross-correlation cost functions. It also provides flags to individually disable specific affine components (shear, scale, rotation, translation) for constrained registration. A secondary source/target pair can be supplied for multi-channel cost evaluation, and a working directory can be specified for intermediate files.

OPTIONS

-verbose: Print progress information during processing.
-clobber: Overwrite existing output files.
-fake: Print the commands that would be executed without running them.
-init_xfm <xfm>: Use the specified transformation as the initial starting estimate.
-noresample: Do not produce a resampled output volume.
-source_mask <mask.mnc>: Apply the given binary mask to the source volume during registration.
-target_mask <mask.mnc>: Apply the given binary mask to the target volume during registration.
-lsq6: Use a 6-parameter rigid-body model.
-lsq7: Use a 7-parameter similarity model.
-lsq9: Use a 9-parameter model. This is the default.
-lsq12: Use a full 12-parameter affine model.
-quaternions: Use quaternion-based rotation parametrisation.
-mi: Use mutual information as the cost function.
-nmi: Use normalised mutual information as the cost function.
-xcorr: Use cross-correlation as the cost function.
-work_dir <dir>: Use the specified directory for intermediate files instead of a temporary directory.
-sec_source <source2.mnc>: Supply a secondary source volume for multi-channel registration.
-sec_target <target2.mnc>: Supply a secondary target volume for multi-channel registration.
-noshear: Disable shear parameters in the affine model.
-noscale: Disable scale parameters in the affine model.
-norot: Disable rotation parameters in the affine model.
-noshift: Disable shift (translation) parameters in the affine model.
-notrans: Equivalent to disabling all translation parameters.
-close: Assume the source and target are already roughly aligned, reducing the search range.

EXAMPLES

bestlinreg_g -nmi subject.mnc model.mnc subject_to_model.xfm

Perform a 12-parameter registration with no shear:

bestlinreg_g -lsq12 -noshear -clobber \
    subject.mnc template.mnc output.xfm output.mnc

Multi-channel registration with a secondary modality:

bestlinreg_g -mi -sec_source t2.mnc -sec_target t2_model.mnc \
    t1.mnc t1_model.mnc output.xfm

AUTHOR

Claude Lepage, Vladimir S. Fonov. McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University.

COPYRIGHTS