Unsupervised Spike Sorting for Large-Scale, High-Density Multielectrode Arrays

Gerrit Hilgen,Martino Sorbaro,Sahar Pirmoradian,Jens-Oliver Muthmann,Ibolya Edit Kepiro,Simona Ullo,Cesar Juarez Ramirez,Albert Puente Encinas,Alessandro Maccione,Luca Berdondini,Vittorio Murino,Diego Sona,Francesca Cella Zanacchi,Evelyne Sernagor,Matthias Helge Hennig

doi:10.1016/j.celrep.2017.02.038

Abstract

We present a method for automated spike sorting for recordings with high-density, large-scale multielectrode arrays. Exploiting the dense sampling of single neurons by multiple electrodes, an efficient, low-dimensional representation of detected spikes consisting of estimated spatial spike locations and dominant spike shape features is exploited for fast and reliable clustering into single units. Millions of events can be sorted in minutes, and the method is parallelized and scales better than quadratically with the number of detected spikes. Performance is demonstrated using recordings with a 4,096-channel array and validated using anatomical imaging, optogeneticstimulation, and model-based quality control. A comparison with semi-automated, shape-based spike sorting exposes significant limitations of conventional methods. Our approach demonstrates that it is feasible to reliably isolate the activity of up to thousands of neurons and that dense, multi-channel probes substantially aid reliable spike sorting.

Highlights

Large-scale, dense probes and arrays and planar multielectrode arrays (MEAs) enable extracellular recordings of thousands of neurons simultaneously (Ballini et al, 2014; Berdondini et al, 2005; Eversmann et al, 2003; Frey et al, 2010; Hutzler et al, 2006; Maccione et al, 2014; Mu€ller et al, 2015; Obien et al, 2015)
If the recording channels are sufficiently well separated, there is no or little overlap between their signals, and spike sorting can be performed by clustering a low-dimensional representation of spike shapes (Harris et al, 2000; Lewicki, 1998; Quiroga et al, 2004)
Spikes are detected using a threshold-based method that exploits dense sampling to improve detection performance and assigns each spike an estimated location based on the barycenter of the spatial signal profile (Muthmann et al, 2015)

Summary

Introduction

Large-scale, dense probes and arrays and planar multielectrode arrays (MEAs) enable extracellular recordings of thousands of neurons simultaneously (Ballini et al, 2014; Berdondini et al, 2005; Eversmann et al, 2003; Frey et al, 2010; Hutzler et al, 2006; Maccione et al, 2014; Mu€ller et al, 2015; Obien et al, 2015). If the recording channels are sufficiently well separated, there is no or little overlap between their signals, and spike sorting can be performed by clustering a low-dimensional representation of spike shapes (Harris et al, 2000; Lewicki, 1998; Quiroga et al, 2004) This approach is inappropriate for dense, large-scale recordings. The size of the datasets makes extensive manual intervention impractical; as much of the process as possible, including quality control, should be automated

Results

Discussion

Conclusion