Extracellular action potentials. As an action potential is fired in a neuron channels open on its membrane, and the electric current of the ions that flow through them and in the resistive extracellular volume around the neuron generates local changes in extracellular voltage that can be sensed with a recording electrode. The figure to the right shows simultenous extracellular and intracellular recordings of a hippocampal neuron under visual control (adapted from Cohen and Miles 2000). The shape of the extracellular potential matches qualitatively the second time derivative of the intracellular potential as predicted by basic cable model, and also varies along the dendritic axis according to temporal distribution of membrane current sources and sinks. Extracellular action potential amplitude is largest close to the soma and decreases with distance to reach typical recording noise level at around 80 microns distance.

“Field potential” generally refer to changes of larger amplitude and longer duration in the extracellular voltage (1-10mV vs. 50-200µV, 10-100ms vs. 1ms). These potentials can be generated by synchronous action potentials in groups of neurons as well as synaptic currents which may often be too small to be detected in the background noise when they occur as isolated events. Spike-waves are a typical example of field potential generated by synchronous firing in groups of neurons.

The processing steps for extracellular signals are detailed in the above figure. Blue and red respectively indicate trace data that come in time-continuous form and event data that come as series of event timing. In Spikoscope main screen access the Extra_Plots settings to display any combination of traces and events. Access Extra_Params settings to set filtering, detection and frequency constants. Access TimeSelection settings or use arrow and zoom buttons to browse through the recording to visually inspect the quality of the processing and analysis.

Filtering. Starting from the source extracellular signals there are 3 main filtering paths plus one test filter. The test filter shows the effect of filter settings only on to the time window currently displayed, instantly revealing the role of the different parameters. The Notch filter options help filter out artifacts like pick-up noise at 50 or 60Hz in case there is any in your recording, while specific Bessel filter settings are required to extract field potential, extracellular action potentials or the most reproducible spike shapes for spike sorting. Once correct settings are determined on the sample time window, you can copy these settings to either of the 3 other filters. Clicking the Filer button will process the whole recording, from beginning to end, for all the extracellular channels, and save the result to disc so that it will be available for the next steps.

Typical filter settings for field potential is a band-pass 0.1-1Hz to 30-100Hz, order 4 Bessel. For extracellular action potential detection the band-pass ranges typically from 50-500Hz to 2500-4000Hz, order 4. For extracellular action potential shapes (for polytrode spike sorting) a high-pass above 200-500Hz is usually good. For field and action potential detection the visual criteria is to have a large uninterrupted rising phase at the center of the events. For action potential shape the criteria is to remove as much low frequency or DC shift as possible to improve consistency of spike shape from one occurrence to the next, while high frequency noise is used by the sorting algorithm (Pouzat et al., 2002).

Detection. From the signal filtered for field potential spike-waves can be detected either on the current time window on the display, to test detection parameters, or alternatively on the whole recording using the Detect button. Same is true for extracellular action potentials (EAP). The detection algorithm is described in Cohen and Miles 2000.

Frequency, synchrony. After the whole recording is processed for event detection, set the frequency fast and slow time constants and kernel shapes. Typically a gaussian kernel is used that replaces each event by a gaussian. For spike-wave the time constants are typically of 5-30s for the fast frequency estimate and 5-10 minutes for the slow frequency curve. For action potentials the time constants are typically 2-10s and  30s-3min. Multi-unit fast and slow frequency curves are then combined to provide indices of single channel locus activity variance as well as covariance among pairs of extracellular channels (Cohen et al., 2006).

Polytrodes. Action potential timing on a set of channels that form a polytrode are merged to define polytrode action potentials. This is done by checking for each extracellular action potential on any channel if there are nearly coincident action potentials on the other channels. If so the timing of the polytrode event is defined as a combination of all of the action potentials involved. The coincidence time window and the way the new timing is determined can be set in the Polytrode_Params settings. Access Polytrode_Plots settings to display any combination of polytrode events and traces.

Polytrode spike sorting proceeds in 4 steps thar are accessed by clicking PolytrodeAutoModel. 1- extract shapes of a sample of polytrode action potentials that will be used to determine the average shape of single units. This is done by extracting bits of trace of the signal filtered for polytrode sorting at the timing of the polytrode action potentials. 2- initialize the model to choose for each polytrode the most significant shape elements. This step also perform some initial calculation that will be later required for automatic sorting algorithm. 3- launch the automatic spike sorting algorithm that will figure out what is the best set of unit shapes that can account for the sample of polytrode action potentials. At this point by clicking EditModel check manually the unit shapes. It is possible to regroup unit shapes in order to compensate for natural spike shape variability, or to remove unit shapes that yield poor reliability. 4- compare action potential with unit shapes to sort all action potentials of all polytrodes in the recording. Inevitably a fraction of polytrode action potentials do not resemble any unit shape and will be tagged as unsorted. Fast and slow time constant frequencies can be estimated for all units by clicking SortedUnitsFreq.

Figures. For specific quantitative or qualitative analysis proceed to the corresponding figure screen by clicking the Figure button. Several figures allow to compare trace and event data, either from distinct extracellular channels, or betwen extracellular channels and intracellular, trigger, or stimulus signals. Check the bibliography for examples.

Declaring extracellular signal channels is done when the recording is opened. The example here shows a screenshot from the File opening screen for a double extracellular recording (chan_8 and chan_9) out of 4 channels. In this example the recording data is actually stored in a directory with one file for each channel. To manually modify channel type, gain, offset or units proceed to the Edit descriptor screen.

Extracellular plots

Extracellular parameters

Spikoscope can analyse electrophysiogy extracellular electrode recordings to reveal distinct components of the extracellular voltage. Spikoscope proceed through a series of step: first some signal channels are declared as coming from extracellular electrodes, then it is possible to bandpass filter these signals to extract slow or fast varying components (which correspond to field potentials and action potentials), then detect the corresponding events, and finally estimate event frequency.

Extracellular recording of an action potential

Spikoscope - Browse and analyze electrophysiology recordings