I noted that there is a bunch of other people using in one way or another anatomical modeling to learn something about the brain. Here is a list of posters and companies I found at the SFN that present work based on anatomical models. This list will be updated throughout the conference:
The next release version of the PAM version of the Waxholm Rat Atlas is already under development. The neural layers (at least stratum pyramidale and granule layer) have volume which makes the neuron distribution a bit more realistic. Computing neuron positions and rendering them at full scale is already possible on average desktop machines. And also the mappings for EC2/3 to DG and CA3/1 works with the volume data.
These renderings show just 5% of all excitatory neurons in the hippocampal formation and for only 20 neurons in EC2 and EC3 their projections to the hippocampus. The axon diameter is at realistic scale (0.75um).
This is just a „small“ test render of the layered version of the Waxholm Rat Atlas. Each main neural layer in the hippocampal formation contains a realistic number of neurons (little pyramids in this case). DG: 1,200,000 neurons; CA3: 250,000; CA2: 30,000; CA1: 360,000. This yielded about 10 Mio. Vertices and it took Blender a couple of minutes to render an image, but overall it was pretty straight forward. A visualization of a full scale model of the rat hippocampus is coming into reach.
The right image also depicts a basal dendritic tree of one single neuron. Now imagine, every neuron has such a dendritic tree. Must be a pretty dense structure.
While I am converting the Waxholm Rat Atlas into a PAM model, I thought it would be fun to see how a single dendrite at realistic scale looks within the entire rat hippocampus. So I imported a model of a dendrite taken from this paper (thanks Corrado) into Blender. As it is impossible in the normal 3d mode to see the entire rat model and the dendrite at the same time, I created a little video, in which I zoom into the dendrite and back. The next step would be to create whole dendritic trees at a realistic scale with the TREEs module.
Only disovered now, that there is a new (February 2015) great resource out there for reconstructing the rat hippocampus as parametric anatomical model. The Waxholm Space Atlas for the rat (and this is the paper about the atlas). The Waxholm Space Atlas contains volume data of the rat brain. Mesh data of the volumes can be obtained from this website. The image above was created based on the mesh data.
With PAM, you can simulate anterograde and retrograde injection-based tracings. Simply position the cursor onto an injection side, select antero- or retrograde tracing in the tracing-panel of PAM modeling, chose an injection radius and then press „Perform tracing“. PAM will mark all post-synaptic (or pre-synaptic) neurons that are affected by the tracer.
Tracing simulation is implemented in a very rudimentary way right now but it already can help to validate the model definition with experimental data. The tracing-feature is already in the master branch.
Cuntz et al. 2010 presents an implementation of the minimal spanning tree algorithm in order to create realistic dendritic morphologies. Patrick Herbers ported this algorithm to Python and wrote a Blender addon for it. The port is now also available in PAM. The dendritic structures in the header have been, for example, created with this addon. A video tutorial shows how to use the addon.
Yesterday, I wrote about a nice Scholarpedia article dealing with axonal conduction delays. The plot of the diameter-to-velocity function is from a paper from 1980 by Waxman. I found it confusing that for an unmyelinated axon with an average diameter of 0.1μm the conduction velocity is predicted to be ~0.7m/s, while experimental evidence indicates that conduction velocity is only about ~0.3m/s.
After some literature search I found that Wen and Chklovskii in 2010 examined this issue and came up with new parameters for the Waxman model of axonal conduction delay.
The most important part is that the coefficient for thin unmyelinated fibers is not ~2.3 as given in the Waxman paper but rather 1.06, in order to explain the experimental data. When it comes to modeling thin unmyelinated fibers (e.g. in the hippocampus), this makes a considerable difference!