Features

Spikoscope has a main display for traces and events, and some controls on the right hand side to 1-select plots and change legends (trace colors, etc.) and 2-set parameters for bandpass filtering, spike detection, etc. Spikoscope handles extracellular channels (as shown above), as well as intracellular (Vm and Ic), trigger and stimulus channels. A selector and channel counter control the display of settings for plots and parameters for each type of signals, for each channel. Setting macros help duplicate settings from current channel to other channels, save and restore settings to/from disk. Clicking buttons on top left will call functions like trace filtering, detection of intracellular or extracellular action potentials, computation of firing frequency or multi-unit activity (MUA), spike sorting, figure generation, correlation analysis, visual field mapping... There are also functions to quantify synchrony in neurons located around one extracellular electrode, or synchrony between neuronal populations at two distinct sites, and functions to detect and estimate the frequency and amplitude of field potentials (Cohen et al 2006). User operations are saved to a journal file to keep track of the analysis process. Larger versions of the above images are available in the screenshot section.

Spikoscope handles tetrodes, or more generally combinations of extracellular recordings (polytrode). Spike sorting is based on an algorithm that maximizes a likelihood estimator, as exposed in Pouzat et al. 2002. Spikoscope includes several implementations of such algorithm: labview implementations of EM (Expectation maximization) and SEM (Stockastic EM), and SpikeOMatic EM/SEM, which is an early C code by Pouzat and Delescluse. The labview implementation has a slight variation: to avoid having too many candidate units converging on the same unit shape, all but one of the units which converge on the same shape is reset to a new random seed every 15 iterations. Another variation: how to guess the number of units? The program searches for more unit shapes than what is actually in the recording, and in a second step the user groups shapes that go together to form a unit, through a model editor. This grouping step also helps cope with natural variation in spike shape and detection jitter. This step is useful to check the set of units (model) on each experiment.