In article <ttrusk.85.2CB2F5DB at its.mcw.edu>, ttrusk at its.mcw.edu (Thomas Trusk, PhD) writes:
|> Any other thoughts on developing such a project?
|>
We're starting to do work on this problem as part of our neuronal database
project (which is primarily a database of timeseries type data, and should
be "publically" available in a year or two). It's actually non-trivial
to do a 3D atlas right; as was pointed out, beyond making sure all of your
slices are properly scaled & aligned, making them look like a brain when
you're all done can be tricky. Some of the labs involved with the database
project would like to be able to use an atlas to do things like identify
specific electrode sites by slicing through the 3D brain & visually
matching the target area with the appropriate region and having the computer
then generate location/name information describing the electrode position
that can be associated in the database with waveform data obtained from that
site. Other labs just want a slick way to fly around some or all of their
brains to get a better feel for anatomy etc. Were trying to develop a
fairly simple system for getting atlas data online from a single (or very
small number of) sectioned brain(s) in a fairly automated way. The
more elegant a mechanism for doing this the better, since we'll be dealing
with at least 4 or 5 different animals from the outset, and we'd like to
end up with something useful to the general neuroanatomical public. We
have a pretty sizable NSF grant for the next few years to do this, and part
of the deal (in accordance with the Human Brain Project I'm told), is
to be able to eventually make the system publically accessible.
I probably wouldn't bother trying to scan/trace data from the P/W atlas
or anything similar. Although it does have a good coordinate system,
their slices don't really have any fiducial points for you to work with,
and since they're from lots of different brains, getting any sort of
registration from slice to slice is going to be a nightmare, even with
lots of good computer assistance.
Anyway, I'd welcome any thoughts/suggestions. We're still sort of in the
planning stages (i.e. we haven't really written much of the code yet), but
we do have a pretty good handle on how to go about this. I haven't really
seen any other serious efforts of this nature around lately unfortunately,
in response to the original poster's question. Marion Ross from NASA Ames
gave a pretty good presentation of their semiautomated 3D reconstruction
techniques a few months ago at the cns93 meeting in DC, but they're dealing
with EM scale structures, not whole brains, and they're interested mainly
in data for finite element modeling of the vestibular macula as I recall.
Her email address is ross at neuron.arc.nasa.gov.
---------------------------
Jeremy Payne
UIUC Neuroscience program /
College of Medicine
jrpayne at uiuc.edu
(217)244-4478
---------------------------