Is there any reliable open-source software for simulating the behaviour of human cortical neurons? I am looking for serious software, so let's assume I have access to a computer with more than 10,000 computing cores.

What I would like to do is to scan a subject's brain with MRI, record EEG during a simple task (e.g. listening to tone pips), project the EEG data to the cortical surface calculated from the MRI (solve the so-called electromagnetic inverse problem), and once that is done try to simulate the behavior of the cortical neurons at the "active" regions.

This simulation part is new to me, the rest I know quite well. All hints and ideas are welcome!

  • 1
    $\begingroup$ I just know NEST, but how would you get the map of the connections ? $\endgroup$
    – agemO
    Feb 27 '15 at 14:02
  • $\begingroup$ The answer by Memming is a bit misleading. People do source-modeling with EEG and MRI all the time in neuroscience, though, as stated, the accuracy is not that great (> 1cm). There's a great variety of algorithms for that, such as dSPM, sLORETA, .. Please mods make this a comment :) $\endgroup$
    – user14676
    Feb 27 '15 at 16:37
  • $\begingroup$ Thanks for the feedback! Yes, of course you can do source localization, assuming there are a few dipoles (or under other simplifying assumptions), but it's difficult to connect to the neuronal level simulation as the OP wants to do. (Sorry. I don't know how to turn your answer into a comment.) $\endgroup$
    – Memming
    Feb 27 '15 at 18:08
  • $\begingroup$ Perhaps I misread OP's question. Upon re-reading, OP only says "simulating the behaviour of human cortical neurons", I guess it could be larger scale model. I'm editing my answer. $\endgroup$
    – Memming
    Feb 27 '15 at 18:25
  • 1
    $\begingroup$ There are a number of neural simulation programs available: Neuron, NCS, Genesis come to mind. However, none of them have anywhere near the computing power to do even a halfway realistic simulation of a human brain. You're looking at roughly 86 billion neurons, with something like 10K synapses each. $\endgroup$
    – jamesqf
    Feb 27 '15 at 18:28
  1. No, not in the neuronal level. Because IMHO this is an ill-posed problem.

EEG combined with MRI won't have the resolution to do solve the inverse problem of neuronal network simulation. There are at least hundreds of thousands of neurons that would be involved in your specific behavioral task, and you have at most a few hundred electrodes. You'll have to make grossly simplifying assumptions about how the brain works to start any inference. You might learn something, but the claim would be somewhat weak.

Personally, I wouldn't work with EEG if you are interested in the mechanistic details of the brain. It is a very useful signal for some cognitive investigations (especially for humans), but the current technology lacks the spatial resolution.

  1. Yes, in the coarse scale network level. (response to @sets' feedback)

If you don't care about micro-scale dynamics of the neural network, and want to model some population-wide activation that changes EEG, as @sets commented, there are various source localization & dynamical system models that you can use. I am not an expert on this, so see @sets' answer, or hopefully someone else will answer in detail.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.