We have used an image registration algorithm from the Computational Morphometry Toolkit (CMTK) to generate bridging registrations between different template brains. During a bridging registration, image data are transformed from their original space into the space of a template brain using both affine and warping transformations.
First, a rigid affine registration translates, rotates and scales the sample brain as a whole to roughly match a chosen template brain.
This is followed by a non-rigid or ‘warping’ registration, where different brain regions are now allowed to move somewhat independently. For this, a grid of uniformly spaced control points is defined over the sample brain, and the space between control points is covered by a B spline interpolation. During the warping registration, individual points in the sample brain are moved independently to match corresponding control points in the reference brain.
To achieve the best match between sample and template brain, the registration algorithm minimises mutual information as a measure of image alignment.
Mirroring registrations follow a similar procedure to bridging registrations. However, rather than transforming data between different template brains, mirroring registrations map data between the two brain hemispheres. To create mirroring registrations, images are first flipped horizontally. The flipped images are then registered to the original template brains using CMTK.