spikesorting

Spikesorting is a computational process used in neuroscience to identify and isolate the electrical signals, or action potentials, generated by individual neurons from extracellular recordings. When electrodes are placed near neurons, they detect the summed electrical activity of multiple cells. Spikesorting aims to disentangle this mixed signal into the distinct waveforms produced by each neuron.

The process typically begins with filtering the raw electrophysiological data to remove noise and isolate the frequency bands where action potentials are most prominent. Then, a threshold is set, and any signal exceeding this threshold is considered a potential spike. These potential spikes are then often subjected to a clustering algorithm. This algorithm groups similar spike waveforms together, assuming that spikes originating from the same neuron will have a characteristic shape.

Once clusters are formed, manual or automated curation is performed to ensure the quality and accuracy of the sorting. This involves visually inspecting the identified clusters, examining their waveform shapes, and assessing metrics like the refractory period (the time after a spike when a neuron cannot fire again) to discard misidentified spikes or merged clusters. The output of spikesorting is a set of spike trains, where each train represents the precise timing of action potentials from a single, identified neuron. This allows researchers to study the activity patterns and functional connectivity of individual neurons within a neural network.