Abstract
Objective. This study examines how the geometrical arrangement of electrodes influences spike sorting efficiency, and attempts to formalise principles for the design of electrode systems enabling optimal spike sorting performance. Approach. The clustering performance of KlustaKwik, a popular toolbox, was evaluated using semi-artificial multi-channel data, generated from a library of real spike waveforms recorded in the CA1 region of mouse Hippocampus in vivo. Main results. Based on spike sorting results under various channel configurations and signal levels, a simple model was established to describe the efficiency of different electrode geometries. Model parameters can be inferred from existing spike waveform recordings, which allowed quantifying both the cooperative effect between channels and the noise dependence of clustering performance. Significance. Based on the model, analytical and numerical results can be derived for the optimal spacing and arrangement of electrodes for one- and two-dimensional electrode systems, targeting specific brain areas.
Highlights
Spike sorting, the identification of individual neurons in extracellular neural recordings, is a fundamental method in neuroscience
We systematically examined the relationship between the geometrical arrangement of recording sites and spike sorting efficiency, and attempted to formalise principles for designing electrode systems with optimal sorting performance
The clusters per channel (CPC) given by the fitted models are shown overlaid on the results of figure 4(b) in solid lines. These results shows that the dual observer model describes the dependency of the CPC yield on the channel numbers used in all the three cases well, including in the lowest signal to noise ratio (SNR) simulation
Summary
The identification of individual neurons in extracellular neural recordings, is a fundamental method in neuroscience. We systematically examined the relationship between the geometrical arrangement of recording sites and spike sorting efficiency, and attempted to formalise principles for designing electrode systems with optimal sorting performance. Takahashi and Sakurai [7] used a multi-wire electrode consisting of 20 wires with the spike sorting software RASICA, and observed the saturation of the well-isolated units as they tested performance on random subsets of channels from 10 to 20 electrodes. They concluded, that 12 channels are enough to reach the plateau of the maximal number of the identified neurons
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have