Ed,
Donna's comments are (as usual) on the mark. I have several additional
questions/comments.
1) What cortical layer do your contours represent? I'm guessing from
your figure that they represent the pial surface, or something close to
it. In that case, there is an inherent problem, because the pial
surfaces from the pre- and post-lunate gyri are in very close
approximation, and any 'jitter' in contour drawing or alignment will
easily result in artifactual 'fusion' across the sulcus.
For this and other reasons, my lab from the outset has advocated the
use of layer 4 (= cortical mid-thickness) as the substrate for surface
reconstructions. Is that feasible for your data sets?
2) Another important issue concerns the spacing between nodes. It is
very important to have approximately equal within-contour and
across-contour spacing. IF you stay with 50 micron spacing between
sections (but see below), you should match that spacing in the contour
drawing process (it is adjustable).
3) I have learned from experience years ago that it can be problematic
to have sections that are more closely spaced than the precision with
which surface points are delineated and aligned. And getting alignment
good to 50 microns from histological sections (with the local
distortions that commonly occur) is HARD! Hence, you might consider
using coarser spacing (100 or 200 microns) for generating the surface
reconstruction. Then if you have data points from sections in between,
they can be entered at intervening levels without the necessity of
drawing the section contours. (BUT then make sure that you number the
contours appropriately - i.e., as even numbers or as every 4th
integer). AND test things out on another trial set before investing a
huge amount of time.
4) What kind(s) of experimental data are associated with your sections?
As Donna indicated, uploading a data set can help with further
diagnosis, but I suspect the above pointers will provide some immediate
redirection.
Hope this helps.
David VE
On Mar 25, 2005, at 1:07 PM, Donna Hanlon wrote:
Hi Ed,
Maybe John and/or David will have better answers, but here's my stab:
1) Smoothing should have helped at least somewhat. You can increase
the smoothing strength and/or iterations, but at some point it should
look smoother.
2) Caret is going to try to make a single surface. It assumes the
slices form a single object, so it will connect them where the bridge
is shortest. I'm not sure how to get around that problem other than
to include enough sections so that they really do join, if that makes
sense.
There's a chance that David/John may want to see your dataset, so if
you could upload your spec file and the files it cites in a zipped or
gzipped tar file, it might be helpful:
http://pulvinar.wustl.edu/cgi-bin/upload.cgi
Donna
On 03/25/2005 12:30 PM, Edward Craft wrote:
Hi,
The attached image is a reconstruction I generated in Caret4.6 from 40
histological slices (50um each, 2mm total depth) taken tangential to
the
cortical surface surrounding the lunate sulcus in the left hemisphere
of a
macaque brain. I have many more slices to trace, but did this trial
reconstruction to see how it was working and encountered two
problems: the
reconstructed surface has a rough, stair-stepped appearance, and
nodes in
the pre-lunate (left-hand side of image) and post-lunate (right-hand
side)
gyri have been linked together.
I didn't find much discussion of topology clean-up in the tutorials,
so I
was hoping you could offer some advice.
(1) How can I make the reconstructed surface fit the section contours
more
smoothly? (Using the Surface: Smoothing option as described in
Tutorial 9
did not seem to help.) Is it possible to generate a surface as smooth
as
that in your macaque atlas from histological slices?
(2) In addition to the separate gyri visible in the attached image, my
sections include buried folds of cortex near the base of the lunate
sulcus.
How can I ensure that the surface topology is reconstructed correctly
when
there are two or more distinct sets of contours, such as these, in a
sequence of slice images?
Any help you can provide will be greatly appreciated.
-- Ed Craft
----------------------------------------------------------------------
--
----------------------------------------------------------------------
--
_______________________________________________
caret-users mailing list
caret-users@brainvis.wustl.edu
http://pulvinar.wustl.edu/mailman/listinfo/caret-users
_______________________________________________
caret-users mailing list
caret-users@brainvis.wustl.edu
http://pulvinar.wustl.edu/mailman/listinfo/caret-users