Realistic Neuron Modeling

[乐琢 Urdro]

Hello everyone. 

We need to build realistic neurons in recent project. Here is the final result:

 

[乐琢 Urdro]

Before to get this, I have tried many ways, such as L-system method:

 

 

Although it can illustrate some important morphology of a neuron, it still not realistic enough.

 

Then I find a website called neuroMorpho.org which contains thousands of neurons morphology data. You can download them for free.

 

These are swc files which contained laboratory data by using 3-D reconstruction method.   The format is very similar to .bgeo or .obj. 

In a swc file, each line represent a point with attributes, such as id, position, type, radius(pscale) and parent. The parent attribute is very important, it defines how these points are connected.

 

For more information about SWC file, see here:

http://www.ncbi.nlm.nih.gov/pubmed/11545695

 

Unfortunately, the SWC files have some mistakes, like z-jump problem. So first we try to fix this problem:

http://www.frontiersin.org/10.3389/conf.fninf.2014.18.00002/event_abstract

[乐琢 Urdro]

Now we have a perfect SWC file. The file shows the "skeleton" of a neuron:

 

 

It is still a "skeleton", it has no "skin"(membrane) and no soma.

 

So then we add membrane and soma to the skeleton. The way to add membrane is very simple, just use polywire SOP, and the soma is base on a sphere modeling. 

 

The really trick thing is the connection between soma and membrane, we called them bridges.

 

 

More information can find here: 

http://link.springer.com/chapter/10.1007%2F978-3-319-09891-3_1

 

[乐琢 Urdro]

Finally we have the perfect neurons. I simply use mantra surface material with falloff opacity to make more like x-ray environment, and some lights.