Plug and Play Spike-Sorting

SEBASTIAN VILLAFAÑE; OSCAR FILEVICH; DANIELA ANDRES

Carloz Paz, Cordoba

Congreso; XXXIV Reunion anual SAN 2019; 2019

Sociedad Argentina de Investigación en Neurociencias

Spike Sorting is the procedure of grouping together action potentials belonging to the same cell, while separating these from action potentials belonging to other cells. As seen from a given extracellular electrode, spikes from different neurons present different shapes, and spikes from one particular neuron roughly retain shape even during very long-term recordings; thus, spike-sorting algorithms rely, in part, on spike shape for classification. Due to the amounts of data collected, spike-sorting by hand becomes impossible for more than a few seconds, and a few electrodes. Even semi-automatic spike-sorting methods, where the final user decides whether a group of spikes does or does not belong to the same neuron are not recommended as they require a human taking decisions which might be biased or operator-dependent. Moreover, spike sorting is part of a chain of tools used to analyze experimental data, or decide the site for clinical deep-brain stimulation electrode implants in Parkinson?s disease. Given the importance of the procedure, a fully automatic, real-time spike sorting algorithm is desired, and some have been proposed. Here we present a spike sorting algorithm that can run in near-real time, based on wavelet encoding of spikes and a genetic algorithm for sorting. The number of clusters is automatically determined (a parameter usually hard to optimize). The performance of this algorithm is compared to a popular spike-sorting algorithm, with good results.