Abstract
The challenge of spike sorting has been addressed by numerous electrophysiological studies. These methods tend to focus on the information conveyed by the high frequencies, but ignore the potentially informative signals at lower frequencies. Activation of Purkinje cells in the cerebellum by input from the climbing fibers results in a large amplitude dendritic spike concurrent with a high frequency burst known as a complex spike. Due to the variability in the high frequency component of complex spikes, previous methods have struggled to sort these complex spikes in an accurate and reliable way. However, complex spikes have a prominent extracellular low frequency signal generated by the input from the climbing fibers. We exploited this to improve complex spike sorting by applying Principal Component Analysis (PCA) on the low frequencies of the signal and show that the low frequency first PC achieves a better separation of the complex spikes from noise. The low frequency data are more effective in detecting events entering into the analysis, and therefore can be harnessed to analyze the data with a larger signal to noise ratio. These two advantages make our method more effective for complex spike sorting. Our characterization of the dendritic low frequency components of complex spikes can be applied in other studies to gain insights into processing in the cerebellum.