PALS Volume Normalization

2005-03-29 (expositions added 2006-09-15), Donna Dierker, Van Essen Lab

We obtained the twelve structural MRI volumes from Randy Buckner's lab in's 711-2C space, cubic 1mm voxels. We registered these volumes and surfaces using five methods:

Click on the links above to see the steps used for each method. The defaults were used as much as possible.

The "space" field in the spec files in CARET_HOME/data_files/fmri_mapping_files specifies which method was used. These spec files list the corresponding fiducial surfaces to be used when specifying Multi-Fiducial Mapping (MFM) in that space. Space is not an ideal term (alternate suggestions are welcome, but "method" is even less informative). Instead of "talairach" or "mni305," these fields contain method-specific tags like "afni," "spm2," "flirt," and "mritotal." But besides 711-2C, there really are only two or three target coordinate "spaces" for PALS_B12:

Our "afni" space is the only true talairach option, although there are other methods that use true talairach space. Both SPM* and flirt (FSL) use avg152T1.mnc, while mritotal (MNI) uses average_305.mnc. (Of course, you can use these tools with alternate targets -- atlas or otherwise, but these are the targets included with the software distributions.) To the best of our understanding (and we've read over and over), avg152T1.mnc and average_305.mnc are in the same coordinate space, but are clearly distinct targets (i.e., one averages 152 subjects, while the other averages 305). They share a common origin and orientation. (This isn't strictly true, since there are right-handed and left-handed versions, but they can be x-flipped as needed to bring them into alignment.)

In our minds, various targets can occupy a common space. For users, for example, the 711-2C and 711-2B spaces are equivalent for the purposes of MFM. The brains that comprise the 711-2C average "fuzzy" volume were first registered to 711-2B. The 711-2C target was developed to better represent normal older brains. The particular atlas target -- optimized to best represent the populations under study, the scanner and acquisition protocol, etc. -- is distinct from the space it occupies. At Washington University, 711-2B, 711-2C, 711-2L, and 711-2Y are all in the same space, but they are distinct atlas targets optimized for various populations or applications/studies. But they're all in the same space, which we call 711-2B. It is also possible to register an atlas target -- say avg152T1 -- to alternate spaces (e.g., true Talairach or 711-2B), and people do things like this to facilitate cross-study comparisons and/or enjoy software benefits that depend on things being in a particular space (e.g., see AFNI's @auto_tlrc usage).

But for the purposes of fMRI mapping expediency, the Caret "space" tag specifies which of the methods above was used to warp the PALS_B12 subjects' fiducial surfaces to talairach, MNI305/MNI152, or 711-2B space.