Spike detection and sorting using PARAFAC2 method.

ABSTRACT

In this contribution we introduce the Parallel Factor 2 (PARAFAC2) analysis as a novel method for the simultaneous detection and classification of neural action potentials. In order to measure these action potentials (spike signals), stem cell derived neuronal cells are cultivated on the surface of a Micro Electrode Array (MEA). Here, the neuronal cells produce ion currents, which can be measured as extracellular electric potentials. Whenever a cell or a group of cells produces ion currents, either spontaneously or evoked by a stimulus, a spike signal can be measured by the electrodes of the MEA. Stimulated cells produce spikes and groups of spikes (bursts) which propagate in space over the MEA. In the recorded data, different source types (e.g., cells which respond directly to external stimuli and cells which are triggered by other neural cells) are characterized by different spike shapes. The proposed PARAFAC2 method is able to separate these spike shapes (sources) in time, frequency and space (channels) enabling an improved performance in noisy scenarios. Furthermore, PARAFAC2 allows for a causality analysis on the measured spike signals (i.e. the identification of different signal paths). Thereby, the PARAFAC2 decomposition is able to exploit the multi-dimensional structure of the MEA data.