Spike_sorting

Due to the high channel count and the need to consider information from many adjacent channels in the sorting of each spike, many existing spike sorting algorithms are insufficient or unusable for data produced by Neuropixels probes.

Automatic spike sorting

Kilosort is an automatic spike sorting algorithm developed for Neuropixels probes that can process the data quickly and accurately. There are currently 5 (1.0, 2.0, 2.5, 3.0 and 4.0) versions of Kilosort, with the latest (4.0) not yet publicly available. There also a Python port based on Kilosort 2.0, which additionally aligns all channels using an FFT-based algorithm.

Find the code for all Kilosort versions here, a paper about how Kilosort 1.0 works here, more information on Kilosort 2.5 here and information about all Kilosort versions here. In 2016, Kilosort 1.0 outperformed other tested algorithms in an objective comparison.

Channel map files for using Neuropixels data with Kilosort can be found here, and a script to generate channel map files can be found here.

Alternatives

Other software that may work well for Neuropixels data includes Spyking Circus and JRCLUST: Janelia Rocket Clust.