Automatic Methods for Motor Intention Recognition from Spike Rates

In this paper we present a method for automatic detection of motor intention from in vivo neuronal recordings in monkeys. The analysis relies on a data base of spike trains collected in a series of experiments aiming to study the hand-eye coordination mechanisms in primates. The neural activity is recorded using a multi-electrode system that can monitor up to fourteen neurons at time. In this work we analyze the possibility to “read” the motor intention from the set of simultaneously recorded spike trains, by combining the information from all the available recordings. We show that the information of interest can be successfully extracted from the data, under some constraints. First, we show the analysis of spike trains, segmented according to the behavioral epochs defined by the experiments protocol, and give the discussion of the proposed method performance in extracting the information of interest, i.e. the presence/absence of motor intention. Also, we consider a less ’controlled’ analysis of entire spike trains, without segmenting, where the relevant information is more mixed with the side-effect processes, and accordingly, more difficult to recognize.