Sign up for Neuralynx Emails



April 2015 TechTip

Speed up offline cluster-cutting using Cheetah and SpikeSort 3D

Anyone who faces the challenge of analyzing electrophysiology data from long recording sessions will know the tedious and slow process to go from raw data to a reliable sample of data.A few tricks can help Neuralynx Cheetah acquisition software users accelerate the transition from raw data collection to sample set analysis.

The first step consists of setting appropriate thresholds within the acquisition entities. Cheetah software allows for each acquisition entity to set filtering, thresholds and input range.

The threshold for the spike detection can be set either in the “Acquisition Entities and Display Properties” by entering numbers for each of the AD channels or in the “Spike Window” by moving the green line on the left side of the waveform plot up or down with the mouse. If the acquisition entity is a tetrode or a stereotrode, independent thresholds can be set for each channel within the acquisition entity.

  • Left: Tetrode Spike Window Right: Acquisition Properties Window

A correct setting of the threshold in each acquisition entity will result in well separated clusters for each of the units present in the single channels, even when two or more units are present on the same AD channel. In order to achieve greater separation for each cluster, it is advisable to check the different distributions based on the several features that can be used for X, Y and Z axis (e.g. Peak, Trough, N-sample, etc.). The multi-dimensional distribution of the clusters will be visualized for a quicker and easier way to represent the neural data.

  • SS3D Feature Space

Use of the Neuralynx Netcom application, Spike Sort 3D (SS3D), allows the user to run cluster-cutting online while the experiment is in progress. Having set the appropriate thresholds, SS3D provides the user the ability to create templates from the waveforms of each cluster. Those templates can then be used within Cheetah to identify spikes online from the different cells within the stream of neural data, providing an invaluable time-saving for offline analysis.