Non linear methods estimating neural network behavior on Micro-Electrode Array technology

In the recent years Neuroscience research is exploiting technologies initially developed only for electronic engineering use: this is the case of Micro-Electrode Array (MEA) technology, where a finite number of channels acquires in vitro neural spiking activity. In this work we present a new method to process time data series from MEA trough an ad-hoc software-framework. Our aim is to build a classifier giving quantitative measures of similarity and statistical dependence among neurons activities recorded in different MEA channels. Methods applied to extract specific information about neuronal behavior are Mutual Information and Dynamic Time Warping. In order to extend the pair-wise information so obtained to the entire neuronal networks on MEA, we have chosen to implement a sub-optimal criterion thanks to Genetic Algorithms (GA): this technique support us to sort MEA channels based on dependent activity, thus providing a global index. We applied it to test the presence of self-synchronization among neurons, which can evolve in time and adapt their self in response to specific external stimuli, such as those of the chemical neuron-inhibitors here analyzed.