Last update: Fri Aug 29 11:54:38 EDT 1997

What am I looking at?

This is the SPIKESORT FAQ. It is a collection of information about spike sorting methods, papers and software, primarily derived from discussions on the SPIKESORT mailing list. In the fine tradition of usenet FAQs, it is arranged as a series of questions and their answers.

Whom is this document for?

The primary intended audience is extracellular neurophysiologists, or those considering doing extracellular neurophysiology. Other neuroscientists may find it interesting but it will probably not be at all useful to those with no interest in experimental neurobiology.

What questions are answered here?

Click on a question to jump to the answer:

• What is spike sorting?
• Why is spike sorting important?
• What makes spike sorting difficult?
• Where can I get software to sort spikes?
• What academic papers are available on spike sorting?
• Which companies provide spike sorting software or hardware?
• How do I join the SPIKESORT mailing list?

What is spike sorting?

Neurophysiologists often record the currents flowing across cell membranes using an insulated wire probe whose exposed tip is placed in the extracellular medium near the soma or other spiking region of a cell. Any such probe will record currents passing across all sorts of physiological membranes close to its tip, but we can approximate this as a small number of close, high-amplitude sources imposed on a background of synaptic, somatic and axonal currents from more distant sources. Spike sorting, then, is the process of identifying the waveforms associated with action potentials in the foreground cell(s) and thereby detecting firing events.

Why is spike sorting important?

The short answer to this is that it is vital for extra-cellular recording from multiple cells. Obviously, sorting spikes from a single electrode can provide signals from more than one cell. But automatic recognition is also essential for any multiple electrode recording as achieving and maintaining an isolation on multiple probes simultaneously can be highly impractical.

What makes spike sorting difficult?

A number of issues make the problem non-trivial.

• The membrane-current waveform produced by a spike in a given cell is not identical each time the cell fires. Local, non-voltage gated conductances may have some effect, but the largest difference is probably due to variations in the inactivation state of voltage-gated sodium channels, for example when one spike follows another very closely. In particular, this may lead to a failure of the dendritic spike, severly changing the extra-cellular current flow.

• The shapes of the foreground action potentials might be quite similar to each other. They are all spikes, after all. The spectral density of the spikes is also similar to that of the background, which is itself partly composed of distant spikes.

• The background activity due to nearby cells and synaptic currents may be correlated to the spiking of the foreground cells.

• The electrode may drift in the tissue during the course of the experiment, changing the current waveform as recorded at the probe tip.

• More than one of the foreground cells may fire simultaneously, leading to a complex combination of foreground currents.

So, how do I do it?

A number of approaches have appeared in the literature over the years. Some of these authors have made software available for general use. Finally, a few companies are selling tools that might be of use.

What academic papers are available on spike sorting?

Here is a list of papers published on the subject:

• M. Abeles and M. H. Goldstein, ``Multispike Train Analysis,'' Proceedings of the IEEE 65(5):762-773 (1977).

• R. Chandra, and L. M. Optican, ``Detection, Classification, and Superposition Resolution of Action Potentials in Multiunit Recordings by an Online Real-Time Neural-Network,'' IEEE Transactions in Biomedical Engineering 44(5):403-412 (1997).

• M. S. Fee, and P. P. Mitra,, ``On the variability of extracellular spike waveforms of cortical neurons,'' Journal of Neurophysiology 76(3):3823-3833 (1996).

• M. S. Fee, P. P. Mitra and D. Kleinfeld, ``Automatic Sorting of Multiple Unit Neuronal Signals in the Presence of Anisotropic and Non-Gaussian Variability,'' Journal of Neuroscience Methods 69:175-188 (1996).

• S. N. Gozani and J. P. Miller, ``Optimal Discrimination and Classification of Neuronal Action-Potential Wave-Forms from Multiunit, Multichannel Recordings using Software-Based Linear Filters,'' IEEE Transactions on Biomedical Engineering 41(4):358-372 (1994).

• C. M. Gray, P. E. Maldonado, M. Wilson and B. McNaughton, ``Tetrodes Markedly Improve the Reliability and Yield of Multiple Single-Unit Isolation from Multiunit Recordings in Cat Striate Cortex,'' Journal of Neuroscience Methods 63(1-2):43-54 (1995).

• M. S. Lewicki, ``Bayesian Modeling and Classification of Neural Signals,'' Neural Computation 6(5):1005-1030 (1994).

• J. S. Oghalai, W. N. Street and W. S. Rhode, ``A Neural Network-Based Spike Discriminator,'' Journal of Neuroscience Methods 54:9-22 (1994).

• W. M. Roberts and D. K. Hartline, ``Separation of multi-unit nerve impulse trains by a multi-channel linear filter algorithm,'' Brain Research 94:141-149 (1975).

• W. M. Roberts, ``Optimal recognition of neuronal waveform,'' Biological Cybernetics 35:73-80 (1979).

• M. Sahani, J. S. Pezaris and R. A. Andersen, ``On the Separation of Signals from Neighboring Cells in Tetrode Recordings,'' to appear in the proceedings of NIPS 97.

• M. Salganicoff, M. F. Sarna, L. Sax and G. L. Gerstein, ``Unsupervised Waveform Classification for Multi-Neuron Recordings: A Real-Time, Software-Based System. {I}. Algorithms and Implementation'', Journal of Neuroscience Methods 25:181-187 (1988).

• M. F. Sarna, P. Gochin, J. Kaltenbach, M. Salganicoff and G. L. Gerstein, ``Unsupervised Waveform Classification for Multi-Neuron Recordings: A Real-Time, Software-Based System. {II}. Performance Comparison to Other Sorters'', Journal of Neuroscience Methods 25:189-196 (1988).

• E. M. Schmidt, ``Computer separation of multi-unit neuroelectric data: A review,'' Journal of Neuroscience Methods, 12:95-111 (1984).

Where can I get software to sort spikes?

Various companies have commercial spike sorting products available. Some academics have also made their software available for free over ftp. Here are the packages that I know of:

• Mike Lewicki's Bayesian spike sorter, sslib, is available for ftp from ftp.etho.caltech.edu in /pub/sslib.
• A manual cluster-based sorter for tetrode data written by Sergei Rebrik called spiker.

Which companies provide spike sorting software or hardware?

• Datawave have both an interface to manual cluster cutting and autocut for automatic clustering.
• Spectrum scientific.
• Alpha-Omega.
• Signal Processing Systems makes a pc-based system for manually-driven template-based classification.

How do I join the SPIKESORT mailing list?

SPIKESORT is a mailing list for developers, and users of spike sorting algorithms as well as any interested on-lookers. To join send a message to spikesort-request@vis.caltech.edu with the one word message body "subscribe".

---