I'll pass this on to Mike Hines (here at NNC) and Brenda Claiborne of
UTSA for further comments. There are several formats for anatomy used
by people who do quantitative morphometry. Eutectics and Neurolucida
have their own proprietary formats, and I am sure there are others.
There is also a format that was created by Rocky Nevin which is used
by Brenda, John Miller @ Berkeley, and others who have their own
quant. morph. system. Translators exist for converting these to a
form that NEURON can handle, and I suppose there are also translators
for Genesis.
> In addition, what tools are available to convert raw voxel data on a cell
> to a cell descriptor file (etc.)?
Steve Senft (here at NNC) created a tool that runs within VoxelMath
(Vital Images, Inc.) which in 15 minutes created the morphometric
table for a cell from a 3D image stack. Badri Roysam's group at
Rochester has done something similar, and I think they're
commercializing it as well.
The usual methods (even with computer assisted morphometry) take
10-12 hours of hard labor.
There are serious limitations with this process, though.
In many if not most mammalian neurons, a big chunk of the surface
area is in the finest processes (< 1um diameter). The latest evidence
indicates that this membrane is electrically close enough to the soma
that it will have a big influence on electrical properties. Errors in
diameter measurement on the order of a half micron or so can cause big
errors in computed RN, not to mention serious distortions of the
electrotonic architecture of the cell.
This means you have to start with hi-res voxel data. Let us suppose
you're dealing with a neuron with a dendritic tree that is contained
in a volume 1000um long and 100um square. For 1um "resolution" you
need 10MB storage--but you really need at least 0.25um resolution, so
you actually need 640MB storage to hold the "raw image stack."
Remember, it comes out of the scanner in grey scale without
segmentation, and before you can separate the cell from surrounding
stuff you need to do a seeded fill on this raw data.
Now we get to the hard part: optics. First, even if your confocal's
objective has a 0.25um lateral resolution, the Z axis is only about
1um resolution. You might be able to work around this by oversampling
and interpolation, or perhaps by deconvolution (heh heh heh, on a
640MB image!) to clean things up a bit.
Second, you can't capture the entire tree in one pass because the
objectives that have sufficient resolution also have small fields of
view. So you have to zipper adjacent image stacks together, fixing
problems of partial overlap & skewness as they arise.
Perhaps a different imaging approach would avoid both the Z
resolution and the field of view problems.
--Ted
===========================================================================
Comment from the moderator:
As pointed out in the reply above, there is no single standard format,
although the Eutectics format is fairly common. Hee are some of the
formats and format converters which I know about:
The Eutectics format is described in J. Capowski, Computer Techniques in
Neuroanatomy, Plenum Press.
Last month, Rogene M. Eichler West (rogene@lenti.med.umn.edu) announced a
beta version of a mesh generator that takes Eutectic files as inputs and
generates GENESIS, NEURON, or custom format output files.
Alex Protopapas at Caltech has written programs to convert both Eutectics and
NED (the format used by John Miller's (of Berkeley) neuron tracing system to
GENESIS. He expects to have them in shape for distribution in about two
weeks, and will make an announcement to comp-neuro at that time.
The GENESIS distribution has a program (NIH_parser.c, in the genesis/src/tools
directory) to convert a format described in Bunow, B., Segev, I. and
Fleshman, J. W. (1985) Biol. Cybern. 53, 41-56. I don't believe that
this format is widely used, however.
The NODUS simulator by Erik De Schutter
(http://bbf-www.uia.ac.be/NODUS/NODUS_index.html) has filters for 5
different formats.
I would welcome descriptions of other format conversion programs, and
announcements of the availability of cell morphology files in various
formats.
Dave Beeman, Moderator of comp-neuro
-------------------------------------------------------------------------