Decrypting Complex Brain Signals

Yury V. Zaytsev from the Simulation Laboratory Neuroscience has developed a new method of analysis for understanding the networks formed by hundreds or even thousands of nerve cells in the brain. The synaptic organization of neuronal networks and neural information processing is the basis of all higher-level cognitive brain functions such as vision or speech. Using multi-electrode array technology, neuronal activity in the form of electric pulses produced during the firing of neurons can be recorded, but these signals do not include any direct information about the topological structure of the associated brain region. Established statistical methods used to reconstruct neuronal networks based on generalized linear models can be used to decode networks consisting of some dozens of neurons. Zaytsev's proposed algorithm is less computationally intensive than the above-mentioned methods and thus allows the reconstruction of networks consisting of thousands of cells. This new approach uses the maximum-likelihood estimation of a minimal model optimized for large networks and reaches an accuracy of more than 99 % under ideal conditions, as has been shown in simulations. The analysis of these larger networks makes such demands on compute resources that it can only be performed on supercomputers. The detailed results are available in the Journal of Computational Neuroscience (http://dx.doi.org/10.1007/s10827-015-0565-5).
(Contact: Prof. Abigail Morrison, slns@fz-juelich.de)

JSC News, 11 August 2015