Neuronal cell spike sorting using signal features extracted by PARAFAC

In this contribution we present first investigations for the signal processing on Micro Electrode Arrays (MEAs) used to observe the growth of biological neural networks. Neural cells produce ion currents, which can be measured extracellularly as differences in electric potentials. In order to measure these potentials (spikes), neuronal stem cells are cultivated on the surface of MEAs. These stem cells are allowed to build interconnections over a period of up to 30 days. If a cell or a group of cells produces currents spontaneously or evoked by a stimulus, a spike signal can be measured by the electrodes of the MEA. However, it remains difficult to identify which cells are signal sources and which cells are simply responding to a stimulus, with the additional complication of identifying the path of the signal itself. An initial solution to some of these difficulties is the use of spike sorting. In general, a spike sorting algorithm can assign spikes to different clusters, from which we can identify the different possible sources in the neural network. Subsequently, causality analysis (i.e. analysis of signal path) can be performed using the clusters. In this paper we present a novel and efficient method for separating neuronal spikes using individual waveform features. Thereby, we use the PARAFAC algorithm for the extraction of the features in order to exploit the multi-dimensional structure of the MEA data.