overlapping spikes

[csn-le]

overlapping spikes

[ferchaure]

Sorry, but Wave_clus doesn't handle overlapping spikes.
If you want to detect both of them you can reduce the value of par.ref_ms (1.5 ms by default). Even though Wave_clus won't sort them correctly.

[jtbirmingham]

Thank you; that was my fear.

I'm just wondering...aren't overlapping spikes, in general, an issue in spike sorting? Ultimately I am trying to sort much busier spike trains (or at least identify the spikes I want), e.g.

Do you have any advice/suggestions on approaches to pursue or research? If I am going to start from scratch I would do something simple (and likely naive) like PCA plus some subtractions and to hope that a simple superposition was at work.

Thank you again!
John

[ferchaure]

Yes, they are a quite an issue. But for some applications, where the overlapping spikes are uncommon, they aren't a priority.

I remember some work about this topic, where it had been checked that the superposition is linear. In theory you could try to decompose each overlapped waveform in a summation of N know waveforms.
But in practice this idea isn't so simple. Have in mind that, in general, any decomposition will have less certainty than a normal sorting.

[jtbirmingham]

Thanks for all your assistance!