mritoself - intra-subject registration of two volumetric data sets mritoself <options> <source.mnc> <target.mnc> <xform.xfm>


Mritoself estimates the linear transformation required to register two volumetric data sets from the same subject (intra-subject registration). The procedure was designed and tested on MRI data although other modalities can be registered as well (but this has not yet been tested).

Mritoself estimates the registration transformation in multiple steps, each successive step refining the previous fit. Currently, the sequence of resamplings, crops, blurs and fits are hard-coded, although there are a number of options to force specific fitting strategies for certain types of data.

Fitting Data

By default, Mritoself uses a mutual information objective function evaluated on the original (unblurred) data volumes. Optionally, the data can be blurred before fitting with the -blur option. If the -gradient option is specified, then the gradient magnitude of the blurred images is used in the fitting process.

Fitting Strategy

A two step fitting process is then applied to calculate the registration transformation. The first begins with an identity transformation, 7.3mm steps, and a medium-small (3mm) simplex to find an reasonable initial transformation. The second (and last) fit is estimated with 4.3mm steps and a small simplex (1.5mm).

If an initial transformation is available, it can be used with the -transform option. This will overide the -identity transform used in the first fit described above.

option : description of option

Generic options

-help: Print summary of command-line options and abort.


Greg Ward and Louis Collins

Copyright (c) 1996 by Greg Ward and Louis Collins