On 10/24/2006 01:58 PM, Graham Wideman wrote:
Donna:
Thanks for your quick reply (and yes, apparently we've crossed paths
on those "other lists" :-). You have indeed alerted us to some places
in the docs that we missed before. And your additional comments were
helpful too.
But needless to say, more questions/comments:
1. Average Area: As I now understand it, this applied to the MFM
procedure, and is the total area above threshold, calculated by
sampling the voxel data (eg: fMRI activation) using each of the 12 B12
surfaces in turn, and then averaging those 12 totals together. Main
use: adjusting thresholds for the MFM procedure to give results
comparable to the AFM procedure. Right?
Comparable, but different in a couple of important respects. The AFM
will give you values that haven't been "messed with" -- i.e., if using
enclosing voxel mapping method, then each node in the PALS_B12 average
fiducial (whichever space you selected -- AFNI, flirt, SPM2, 711-2C)
will be assigned the intensity value of the voxel enclosing it. No
averaging happens at the volume level. The averaging happens at the
surface coordinate level and has been packaged in your caret release (in
caret/data_files/fmri_mapping_files).
MFM is comparable in the sense that both AFM and MFM are for mapping
group results onto a multi-subject surface atlas, but MFM intersects
each of the 12 PALS subjects surfaces with the volume, and then averages
in surface-land. (And if you threshold the volume going into the mapper,
then it will compute the suprathreshold surface area for each subject,
to generate an average suprathreshold activation area, and set the
average area threshold such that this much area is supra-threshold.) So
your data values get massaged -- clearly not what you want for ROI type
volumes, and often not what you want for your stats. But it may better
represent the potential extent of your blob and the uncertainties with
this method.
2. Mid-cortex surfaces, and your FS-related suggestion: Where we use
your provided surfaces to sample (map) the fMRI data, we think we
should be using mid-cortical or pial surfaces, but the surfaces that
the tutorial leads us to appear to be grey/white boundary surfaces.
Are we thinking straight on this, and are there mid-cortical versions
of PALS-B12 in AFNI+tlrc space, for example?
All the PALS_B12 surfaces are mid-cortical thickness surfaces. (In fact,
Caret doesn't generate GM/WM and pial surfaces; it only generates a
midthickness aka "fiducial" surface.) If you average your Freesurfer
pial and white surfaces, they'll look remarkably similar to Caret
fiducial surfaces. You won't see razor sharp gyri on our surfaces.
Yes, there are PALS_B12 surfaces in AFNI+tlrc space. Note that the
registration was done before @auto_tlrc, using the marker method in
AFNI. Details on the space flavors and how they got that way are here:
http://brainvis.wustl.edu/help/pals_volume_normalization/
3. "volume-averaged group data": This phrase shows up in a few places,
and we're puzzling over whether it means something different or
special compared to what we *think* it means. Eg: on p67 of
Caret_Tutorial_Sep22.pdf:
"Many fMRI studies are analyzed by registering the MRI volumes from
individual subjects to a standard stereotaxic atlas, then carrying out
statistical analyses on the volume-averaged group data."
I think the phrase "carry out statistical analysis on volume-averaged
group data" possibly refers to "voxelwise statistics, using subject
variables for grouping"? Eg: 256x256x256 voxels by 10 subjects go in,
256x256x256 voxels-worth of means, t, z or whatever come out.
This is what we mean. "Voxelwise statistical output maps for group
studies" is probably a better term than "volume-averaged group data."
You've done your intersubject analysis in volume-land and you want to
visualize it on a surface, but aren't sufficiently motivated to
reconstruct your individual subjects' surfaces; then map it to PALS_B12.
Or is there something more to it? If it's a different concept, what is
the "group data" that is being averaged over a volume, and then what
is the statistical analysis that is performed upon it?
No -- this is it.
Thanks,
Graham
_______________________________________________
caret-users mailing list
caret-users@brainvis.wustl.edu
http://pulvinar.wustl.edu/mailman/listinfo/caret-users
--
Donna L. Dierker
(Formerly Donna Hanlon; no change in marital status -- see
http://home.att.net/~donna.hanlon for details.)