A spike-sorting method based on hierarchical clustering to discriminate extracellularly recorded simple spikes and complex spikes from cerebellar Purkinje cells
Background: In the cerebellar cortex, Purkinje cells are the only output neurons and exhibit two types of discharge. Most Purkinje cell discharges are simple spikes, which are commonly appearing action potentials exhibiting a rich variety of firing patterns with a rate of up to 400 Hz. More infrequent discharges are complex spikes, which consist of a short burst of impulses accompanied by a massive increase in dendritic Ca2+ with a firing rate of around 1 Hz. The discrimination of these spikes in extracellular single-unit recordings is not always straightforward, as their waveforms vary depending on recording conditions and intrinsic fluctuations. New Method: To discriminate complex spikes from simple spikes in the extracellular single-unit data, we developed a semiautomatic spike-sorting method based on divisive hierarchical clustering. Results: Quantitative evaluation using parallel in vivo two-photon Ca2+ imaging of Purkinje cell dendrites indicated that 96.6% of the complex spikes were detected using our spike-sorting method from extracellular single-unit recordings obtained from anesthetized mice. Comparison with Existing Method(s): No reports have conducted a quantitative evaluation of spike-sorting algorithms used for the classification of extracellular spikes recorded from cerebellar Purkinje cells. Conclusions: Our method could be expected to contribute to research in information processing in the cerebellar cortex and the development of a fully automatic spike-sorting algorithm by providing ground-truth data useful for deep learning.