Formation of synchronized neural assemblies

Assemblies of (transiently) synchronized neurons have been observed in a variety of neural systems and are thought to encode sensory information or store short-term memories. This phenomenon is reminiscent of the formation of clusters in models of coupled phase oscillators. In fact, the dynamics of neural networks can be described by systems of coupled phase oscillators as long as networks have weak connections and the neurons’ firing frequencies are roughly constant.

We have investigated the mechanisms leading to the emergence of these neural assemblies with models of coupled oscillators. In particular, we used experimentally estimated phase-resetting curves of real neurons (mitral cells) to realistically describe neurons as phase oscillators. The simulations revealed that the formation of synchronized assemblies is a rather general phenomenon. In fact, our mathematical analysis yields a necessary condition for synchronized neural assemblies to occur (R.F.Galán et al., Neurocomputing, 69(10-12), p.1112, [PDF]).

In our studies we found that i) the estimated PRC of principal cells in the olfactory bulb, the mitral cells, have partially positive and partially negative domains suggesting that they behave as resonators or type II neurons; ii) these PRCs possess higher order harmonics, a necessary condition for the formation of more than one oscillator clusters (synchronized neural assemblies); iii) inhibitory interactions between mitral cells, as in the olfactory bulb through granule-cell-mediated inhibition, lead to the formation of a dominant synchronized assembly; iv) fluctuating inputs (or noise) into oscillator networks induce transitions between assemblies in a way that is reminiscent of a “winnerless competition” dynamics.

Figure: Raster plots showing the dynamical formation of synchronized neural assemblies in the presence of background noise and heterogeneous synaptic strengths. An excitatory network converges to three equidistant clusters. An inhibitory network converges to two close assemblies (one assembly contains only two neurons). Neural traces were reordered to make the assemblies more apparent. The transition from uniformly distributed initial phases to synchronized assemblies is fast (occurs within 1 to 3 natural periods).


Back to research webpage

Last update: May 2009