Precise 3D morphology

Managing anatomically complex model cells with the CellBuilder. Importing morphometric data with NEURON's Import3D tool or Robert Cannon's CVAPP. Where to find detailed morphometric data.
Post Reply
ziemek
Posts: 34
Joined: Thu May 23, 2019 8:02 am
Location: Warsaw, Poland
Contact:

Precise 3D morphology

Post by ziemek » Tue Dec 10, 2019 2:41 pm

Is there any way to define precise 3D morphology based on vertex/triangles in x,y,z space?

SWC format as well as hoc's pt3dadd() function define cylindrical compartments, which in case of small morphological elements (like precise spine geometry) may differ too much from the needs.

There is SWC Plus specification (https://neuroinformatics.nl/swcPlus) which allows much more control over geometry.

Is there any way to load SWC Plus to the NEURON? Or do you know any other way to define precise 3D geometry?

ted
Site Admin
Posts: 5689
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Precise 3D morphology

Post by ted » Wed Dec 11, 2019 12:00 am

SWC Plus doesn't really seem to exist. Its github site shows one contributor and three commits, the last commit being in 2016. Has there been a single publication about SWC Plus, ever? Who uses the SWC Plus format? Are there any tools that generate or read SWC Plus files? Are there any morphometric data files that employ the SWC Plus format? Of course, anyone with the interest, skills, and time, is free to develop whatever they like, but something that isn't used by anyone isn't a very attractive target.
in case of small morphological elements (like precise spine geometry)
This is starting to enter the domain of MCell and Virtual Cell. How do they deal with space?

ziemek
Posts: 34
Joined: Thu May 23, 2019 8:02 am
Location: Warsaw, Poland
Contact:

Re: Precise 3D morphology

Post by ziemek » Wed Dec 11, 2019 5:23 am

Yes, I see that swcPlus is not adopted, but I understand their manifesto - SWC is quite crude. Since the NEURON allows RxD and 3D mesh with small dx cubical compartments, it's reasonable to ask - how to model the cell with precision much greater then cylindrical approximation?

I know that the NEURON supports Neurolucida format. Does it approximate compartments to cylinders as well? Or is there any other way to model in the NEURON with higher precision than cylinders?

And a question to you but also the community as well - don't you feet that cylindrical approximation left too much information outside? It maybe irrelevant for the big network, but for a bunch of neurons or especially for a single cell model with dendritic spines - it maybe relevant. What's your opinion?

ted
Site Admin
Posts: 5689
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Precise 3D morphology

Post by ted » Wed Dec 11, 2019 11:05 am

Fine spatial detail doesn't matter much for electrical signals, which have length constants on the order of tens or hundreds of microns. It can be important for chemical signals, especially intracellular calcium and second messengers, not to mention ions and neurotransmitters in tissue extracellular space--and so I refer your question to the developers of NEURON's rxd module.

ramcdougal
Posts: 184
Joined: Fri Nov 28, 2008 3:38 pm
Location: Yale School of Public Health

Re: Precise 3D morphology

Post by ramcdougal » Wed Dec 11, 2019 12:09 pm

NEURON doesn't use cylinders.

It uses truncated cones ("frusta of a right circular cone") for the electrophysiology.

For 3D reaction-diffusion domains, we use essentially the algorithm of McDougal et al., 2013 to convert Neurolucidia-style point-diameter measurements into watertight surfaces. The SWC-plus data exported by HBP Neuron Morphology Viewer presumably uses a conceptually similar heuristic for inferring consistent triangular meshes. Either way, the constructed surface is simply that... a construction.

There is, as yet, no ability to load SWC-plus, STL, or anything like that for a reaction-diffusion simulation, HOWEVER the rxd module is constructed in a modular way and one could in principle do this without modifying any existing NEURON code by defining an appropriate Geometry class... see for example the definition of inside in geometry.py, which simply directs voxelization issues to geometry3d.voxelize2.

ziemek
Posts: 34
Joined: Thu May 23, 2019 8:02 am
Location: Warsaw, Poland
Contact:

Re: Precise 3D morphology

Post by ziemek » Thu Dec 12, 2019 5:25 am

Thanks for a very precise answer. I will go through references, but in the case of Neurolucida (which I'm actually not familiar with) - does it use as well truncated cones architecture? I'm asking because whenever I see its reconstructions with spines, it seems to be very precise.

ted
Site Admin
Posts: 5689
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Precise 3D morphology

Post by ted » Thu Dec 12, 2019 10:49 am

it seems to be very precise
What seems to be very precise? Spine location, detailed shape of spine heads, what?

The people at MicroBrightField are very helpful. Address questions about Neurolucida to them and they'll be glad to tell you what you want to know.

Or search the Internet for
neurolucida file format
and download NL9.pdf, which is the user's guide for Neurolucida version 9. Read the relevant entries (hint: "neuron tracing") in that, especially chapter 9.

Or talk to someone who has used Neurolucida.

Or do all 3.

Post Reply