I would like to use SPM8/Dartel with segmented tissue class
images obtained from another software program to
register/normalize to MNI space.
As I understand it, Dartel can take segmented tissue class
images from either Segmentation (c* images, in "native" space,
and requires normalization parameter output file) or New
Segmentation (rc* files).
I would like to emulate using rc* images as input to Dartel; a
couple of questions:
1. What does the "r" in rc* stand for? Are the images already
registered, or do are the parameters saved in the "header", and
what template(s) are/were they registered to?
The "r" stands for resliced. The images are brought into the closest
rigid-body alignment with the tissue probability maps released with SPM.
This involves taking the parts of the nonlinear deformation fields
estimated by the segmentation that relate to grey matter, and fitting a
rigid-body transform to them using a Procrustes analysis. Note also
that there is a second matrix in the headers that encodes a relative
mapping (via the q-form fields) between the "imported" and
"native-space" images.
2. Are the segmented tissue class images binary, probability, or
some other format?
They are supposed to encode tissue probabilities, and have values
between zero and one. Binary images should also work for DARTEL though,
providing that the values are either zero or one after accounting for
scalefactors in the headers. Note that Dartel also works with another
"background" class, that it creates by subtracting (for example) the rc1
and rc2 data from 1.0.
Best regards,
-John
--
John Ashburner <[log in to unmask]>
|