Spike (action potential) detection and sorting.

Bruce Land and Andrew Spence


Neurobiologists often want to detect and characterize action potentials. For this journal club we decided to talk about techniques for finding spikes in noise (random and 60 Hz) and techniques for separating the sources of muliple spikes.

There are two basic questions:

The talk broke into two parts. Bruce covered detection and Andrew covers sorting.


Generally you need to do the following to find spikes:

Bruce decided to implement some of these techniques in Matlab to see how they performed:


Spike sorting of a multiaxon system gets hard becuase of overlaps in time between several APs. The sorting scheme that Andrew talked about attempts to separate out overlapped APs (reference 4). The scheme has several steps:

  1. Record from several electrodes. More electrodes means more total data to aid in AP separation.
  2. Construct AP templates (often by hand) from data that appears not to be overlapped. Andrew did this by isolating "nice" looking APs, then performing principle-components analysis to look for clusters, then using cluster members to make the template.
  3. Using each template, construct optimal filters, which at any frequency have an amplitude proportional to the energy in the target template, divided by the sum of the energy in the noise plus all other templates.
  4. Pass each recorded spike train through the optimal filters to (hopefully) isolate each axon.

He did his talk as a slide show.


  1. "Neural Spike Sorting Under Nearly 0 dB Signal-to-Noise Ratio Using Nonlinear Energy Operator and Artificial Neural Network Classifier," K.H.Kim and S.J.Kim, Biomedical Engineering, IEEE Transactions on , Volume: 47 Issue: 10 , Oct. 2000 Page(s): 1406 -1411
  2. “A new interpretation of nonlinear energy operator and its efficacy in spike detection,” S. Mukhopadhyay and G. C. Ray, IEEE Trans. Biomed. Eng., vol. 45, pp. 180–187, Feb. 1998.
  3. Signal Processing for the Multiunit Extracellular Neural Signal Recording with Low Signal-to-Noise Ratio
  4. Gozani, S. N. and J. P. Miller (1994). Optimal discrimination and classification of neuronal action potential waveforms from multiunit, multichannel recordings using software-based linear filters. IEEE Trans. Biomed. Eng. 41, 358--372.
  5. Spike sorting FAQ