Thalamocortical loops, multi-compartment neurons and more...

SpikeFun, improved...

SpikeFun v0.46 and v0.47 contain significant improvements in the simulation modelling:

First and the biggest addition are multi-compartment neuron models (as per Izhikevich, 2007). At the moment, only pyramidal neurons have multiple dendritic compartments (used for modeling of the apical dendrites), but this will be gradually extended to all other relevant neuronal types. Single-compartment model is still used exclusively in the "box/cuboid" simple geometries (which can be observed by the higher simulation speed despite the same number of neurons).

Second big change is the simulation of the thalamocortical system (try the CortexDemo_xxx.bat files supplied with SpikeFun demo) - it comes with the several interesting features:

  • 6-layer cortical anatomy
  • Model of the thalamus and thalamocortical loops
  • Models of pyramidal, spiny-stellate, basket, non-basket, thalamic relay and reticular thalamic neurons
  • Neuronal connectivity modeled upon V1 visual cortex of the cat (as per Binzegger et al.)
  • Simulation of fMRI BOLD ('F' key)
  • Simulation of intra-cortical EEG (icEEG is using the area around the selected neuron)
  • Optional insertion of miniature post-synaptic potentials (mPSPs or 'minis') to simulate neuronal noise

Thalamocortical simulations can be either configured in the configuration options (beware: many internal parameters are still not exported to the "configuration" page + many options in the configuration page are obsolete for the thalamocortical system simulation), or you can use one of the CortexDemo_xxx.bat presets - CortexDemo_Huge script will simulate 1 million neurons with 200 million synapses (and will require 22-32 GB of RAM for doing so).

In addition, stimuli insertion ('L' key) operates by inserting current in the specific-thalamic neurons.

TODO: Axons are still very simple, and the pending work is adding axonal bifurcations and dendritic compartments for other neuron types apart from Pyramidal neurons. There are also many bugs still :( For example, neuron picking is still not working correctly with non-box/cuboid simulaitons.

Following picture shows SpikeFun 0.47 running a thalamocortical simulation and with selected multi-compartmental pyramid neuron (of cortical layer III). Membrane-potential / Synaptic conductance graphs are shown for one dendritic compartment and dendritic spikes are clearly visible: