Principles of Optimal Electrode Geometry for Spike Sorting
01/30/2020
Zoltán Somogyvári3, Róbert Tóth1, Viktor Varga2
1 Department of Biomedical Engineering, Imperial College of Science, Technology and Medicine, London, United Kingdom
2 Department of Cellular and Network Neurobiology, Institute for Experimental Medicine, Hungarian Academy of Sciences, Budapest, Hungary
3 Complex Systems and Computational Neuroscience Group, Wigner Research Centre for Physics, Hungarian Academy of Sciences, Budapest, Hungary
Performing spike sorting on extracellular multi-electrode recordings is the most widely used technique to track the temporal activity of high numbers of neurons in vivo, both in anaesthetised and awake animals. Despite the significant effort invested in designing high channel count and configurable electrode systems, to date no principles have been determined on how the geometrical arrangement of electrodes influences the number of high-quality spike clusters obtained through this process, or how electrode contact points should be arranged to achieve an optimal number of successfully identified neurons. The present study examines the factors through which the geometrical arrangement of electrodes influences spike sorting precision, and consequently attempts to formalise principles for the design of electrode systems enabling optimal spike sorting performance.We performed simulations, running automatic clustering algorithms on artificial data series using spatio-temporal waveforms recorded in vivo as templates, in which the known spike-times made it possible to evaluate the quality of the resulting clusters from spike sorting. The effects of amplitude ratios in multichannel electrode recordings on the quality of spike sorting were tested. Using waveforms measured in vivo, we showed that high signal-to-noise ratio (SNR) on one channel is not sufficient to reach higher cell/electrode count. The maximal yield of high quality cluster per channel was reached when the SNR exceeded a certain threshold on at least two channels. Based on this result, the following design principle for optimal electrode geometry was formulated: in order to maximise the yield of high quality clusters, the given M electrode channels should be arranged so that the volume observed by at least two electrodes is maximal. Using a simplified isotropic approximation for the micro-field potential generated by single action potentials, analytic results were derived for optimal spacing and arrangement of electrodes for one- and two-dimensional electrode systems.