On 11/28/2006 03:32 AM, Andrew Reid wrote:
Question about documentation:
Does a documentation file exist that outlines each of the various file
types used in Caret, including:
What sort of information is being represented by the data
How this data is used by Caret
How the file types interact (e.g., how area colour files are used by
probabilistic atlas files, etc.)
File formats:
http://brainmap.wustl.edu/caret/caret5.5_help/file_formats/file_formats.html
File types:
http://brainmap.wustl.edu/caret/caret5.5_help/file_formats/files.html
What sort of info is represented by the data: Probably the best doc for
this is the 9/2006 tutorial:
CARET_TUTORIAL_SEPT-06
http://sumsdb.wustl.edu/sums/directory.do?id=6585200
A quick summary:
paint = discrete/categorical/ROI attributes by node
metric = scalar for each node, typically functional -- used often as
"overlay"
surface_shape = scalar for each node, typically anatomical -- used often
as "underlay"
Note: metric and surface_shape formats are identical, so you can open
one file as the other type.
Areacolor maps paint names to colors; palette defines same for metric/shape.
?
Basically, I'm looking for a straightforward description of the
application in terms of its data model, its I/O model, and how it is
organized. The tutorials are great for specific purposes, but in my
case I'd much rather have an understanding of the application and
figure out the tools based upon this.
Sorry, nothing like the ITK software developer's guide, if that's what
you're looking for.
Specifically, I would like to be able to compare a delineated lesion -
registered to a template - to a probabilistic atlas, in order to get a
set of nodes in the PALS surface representing the lesion extents
within the cortical sheet, with each node pointing to a probability
value of being situated in a specific anatomical structure. This way I
can get an idea of which structures are being disrupted by the lesion,
and with what probability.
This seems doable, provided you segment your lesioned brain and register
it to PALS_B12, which may prove challenging. We're grappling with
similar applications on our end. I'm not sure how far anyone has gotten
with this on sizable lesions. Anyone who has done this is encouraged to
share your experience.
I've been playing with the tutorials and various datasets, but can't
seem to get an understanding of how these probabilistic atlas files
work, and how to go from my delineated lesion to a probabilistic
mapping of this lesion.
So you could, in theory, map a ROI paint volume of your lesion onto the
PALS_B12 average fiducial surface in the same space as your ROI. Or map
it as a functional volume, if that proves problematic. Then use
Surface: ROI to threshold the resulting metric (or just select the
paint, if mapped as ROI paint) and do generate report on selected nodes,
with the probabilistic atlas surface loaded. I'm not sure exactly how
this would work, or if the result would be sensible, but have you tried
something like this?
Thanks in advance!
Andrew
--
Donna L. Dierker
(Formerly Donna Hanlon; no change in marital status -- see
http://home.att.net/~donna.hanlon for details.)
