I was wondering whether anybody has experience using interspike-interval histograms for spike sorting and evaluating single unit definitions. Are there studies showing typical ISI histograms of single units in rodent barrel cortex? Is it possible to use ISI histograms to identify single/multi-units or a certain cell type (e.g. based on size or width of the peaks, or number of peaks)?
Hi, I guess the following:
J Neurophysiol. 2004 Jul;92(1):600-8. Epub 2004 Mar 31.
Characterization of neocortical principal cells and interneurons by network interactions and extracellular features. Barthó P, Hirase H, Monconduit L, Zugaro M, Harris KD, Buzsáki G.
is a good starting point.
Hope it helps.
The most important when evaluating spike sorting method and single-unit definition is to check whether your units display the expected refractory period. Unless you're recording from fats spiking neurons, you should expect 2-3 ms refractory periods, so the ISI should all be >2ms at least.
E.g. this should be the case for pyramidal neurons.
The ISI histogram may provide some information about rhythmical properties from your isolated spikes, i.e., whether a neuron show a rhythmic pattern or perhaps a bursting mode of spike trains. In some occasion this information may characterize specific neurons. As an example, VTA dopaminergic neurons show a typical bimodal ISI, where the shorter mode represents the intraburst frequency whereas the higher mode corresponds to the interburst interval. These values could indicate the presence of spikes of a precise neuron, but this information is not sufficiently robust.
The more precise electrophysiological characterization of a neuron implies a good spike sorting method (based in Principal component analysis) and a exact metric of the spike waveform. ISI histograms can be considered an initial information previous to a more accurate methods. The previous answer of A. Leblois is a good example of this starting point of analysis with ISI histograms.
Thank you very much! I really appreciate your helpful comments and literature suggestions.
Tou can use interspike interval to get the frequency of your spikes. A discrimination can be done on the basis of pek amplitude or rise and decay time, for istnce by ysing the mini analysis program.
I want to second Arthur's suggestion that simply checking if you have any ISIs < 3 ms will tell you if you still have several units in your cluster.
Claudia,
Matthieu Delescluse, Pascal Viot, Jean Diebolt and myself showed a while ago how to include ISI information during spike sorting (Improved spike-sorting by modeling firing statistics and burst-dependent spike amplitude attenuation: a Markov chain Monte Carlo approach Journal of Neurophysiology 2009, 91: 2910-2928)... it comes at a fairly high computational cost. I have few ideas on how to speed things up, but since I didn't work with data requiring this kind of sophistication lately, I did not push things further. If anyone is interested in collaborating on that, I'm ready to start working on it.
Christophe
PS: there also a paper by Fee, Mitra and Kleinfeld (1996) J Neurosci Methods 69: 175-188 where ISI information is included in the sorting algorithm. It's very different from what we did.
- Badges
- Science topic
- Similar topics
- Psychology