Import3D's treatment of dendrite connections to soma

[FRB121]

Originally, we had trouble having import 3D connect primary somas properly. we traced the soma as a single oval and tried to trace the dendrites so that they will began within the boundaries of the soma so import3D will recognize this is where the dendrite should connect to the soma. Most dendrites seem to connect properly, however a few dendrites were recognized as connecting to the soma. How does import3D decide to connect the dendrites to the soma? Are we correct in assuming that the process need to beginwithin the boundaries of the soma in order to connect to it? If so, how far within the boundaries should the dendrites begin? If we do extend the dendrites into the soma, would that be adding extra length to the dendrites that does not reflect the actual dimensions of the cell? The two options we see are:
1. to not extend the dendrites into the soma, hope for the best, and fix errors in the cell morphology by hand (which we have not been able to do, any advice?)
2. to start each process with a dummy point at the center of the soma and place the second point of the dendrite at its actual beginning, and after using import3D, deleting the extra starting point manually.

Thank you so much for your help. Any advice will be greatly appreciated!

[hines]

we had trouble having import 3D connect primary somas properly

If there is only one soma, every dendrite should get connected to it. But an
informational message will be printed for each dendrite whose point does not lie
inside or on the boundary.

however a few dendrites were recognized as connecting to the soma.

Meaning they were not supposed to be connected to the soma but the end of
a nearby dendrite? Check to see if the z positions are such that the ends are still
near each other.
It should not ever be necessary to artifiicially add points.
Please send me the original raw data file so I can see what is going wrong. Send
as a zip file to michael.hines@yale.edu.

[ted]

FRB121 sent the raw data to hines, who examined it and sent FRB121 the following
messages via email:

4/28/2008 8:31 PM
The problem is that the points starting at 4242-4291 are reversed.
Hence 4242 which is actually the most distal point
on the main dendrite is logically closest to the soma and so is
connected by a long virtual wire.
In fact, it looks to me like every point on the main dendrite (but not
the points on the branches off of it) are numbered from distal to
proximal. (you can see this by selecting one of the points and then
incrementing the line number and seeing the selection point move toward
the proximal end)

That seems like a bizarre convention to me (tracing from distal to
proximal). Is it at all common? NEURON is flexible enough to handle
it. Disconnect the distal section (it is connected at arc position 0)
and connect the proximal section
at arc position 1 to the soma.

I also see that you defeated my heroic attempt at analyzing the
18 contour soma stack by having all the outlines entirely identical
so that it really is not a 3-d shape at all. I admit I never thought
of that case during the programming. For now, why don't you just
delete 17 of them from the file

4/28/2008 8:55 PM
I exported the original to neuron and see that the proximal
to distal path sections are
dend[32] // should be connected to soma
dend[24]
dend[8]
dend[7]
...
dend[0]

so the fix (if you did no want to change the ASC file) would be
objref rvec
rvec = new Vector()
rvec.append(0, 1, 2, 3, 4, 5, 6, 7, 8, 24, 32)
for i=0, rvec.size-1 dend[rvec.x] disconnect()
for i=1, rvec.size-1 connect dend[rvec.x](1), dend[rvec.x[i-1]](0)
connect dend[0](1), soma(.5)

but check the shape and topology() carefully in original and
transformed to make sure
the transformation is correct.