small amplitude cluster blocking detection of another?

[tens.r...@gmail.com]

Hello everyone,

I've noticed an interesting problem that I wonder if tweaking the params might help with. I am spike sorting data off a dense laminar array. For very large amplitude units, I encounter instances where the spike sorting software seems to miss large amplitude spikes when they happen to collide with a smaller one. In the example attached, (twoclus.png) spyking circus is attempting to detect a small amplitude cluster and a large amplitude cluster (blue and red). In instances in which they occur closely together in time, it sometimes fails to classify the spike on large clustered unit. In the second image (error.png) you see an instance where spyking circus actually misses a large amplitude unit by classifying that spike as solely belonging to the small amplitude unit cell. The waveform view and inspecting the traces suggests this happens often. Any suggestions as to how to deal with it. I wonder if it is something in the template matching procedure that can be tweaked? Any help would be greatly appreciated.

Thank you,

Manu Raghavan

[Pierre Yger]

Dear Manu

Which version of the software are you using? Latest? It is hard to see on your png plots, but the missed big spike in the "error.png" file does not look like the one that is fitted, in red, in the previous image. Is it? Some nearby channels are not displaying the same information, therefore a miss is likely here. However, let me know what you mean when you say "often". Can you gather few snippets of missed spikes, and look at the averaged waveforms?

Best

Pierre

--
You received this message because you are subscribed to the Google Groups "SpyKING CIRCUS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spyking-circus-u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/spyking-circus-users/d9c24734-34a1-4607-89bc-786e24f746c2n%40googlegroups.com.

[tens.r...@gmail.com]

Hi Pierre,

>> It is hard to see on your png plots, but the missed big spike in the "error.png" file does not look like the one that is fitted, in red, in the previous image. Is it?

No the missed spike in error.png is a spike that occured shortly after the spike fitted in red in the previous image.

>> Which version of the software are you using? Latest?

   I am using the latest version of the software, I installed it just last week. I can update it and rerun if you've pushed new changes since then.

>> However, let me know what you mean when you say "often". Can you gather few snippets of missed spikes, and look at the averaged waveforms?

By eye (and other spike sorting algorithms observed using spikeinterface) I have identified at least 800 instances in which this happens. I gathered snippets of time around 300 of these spikes (+/- 1ms before/after) and display them. Individual spikes are drawn with gray lines, the mean waveform associated with them in thick black. I plot for just the relevant channels. I also plot 300 spikes identified by spyking circus belonging to the red cluster previously shown and plot their mean waveform in red, and finally I plot the overlap in mean waveform between missed and identified spikes. Is this what you were looking for? I see that on channel 23 there is a tendency for the mean waveform to differ. Is this what spyking circus is sensitive to?

Manu