Lupin clustering distant spikes

[aidapiccato]

I'm using Lupin to spike sort, and have found that it's working well with the exception that it tends to cluster together very distant spikes. Here's an example:

The lower right plot shows peak y locations over time - it's incorporating spikes that are up to 2000 microns distance apart into the same cluster. Are there parameters I can adjust to avoid this?

[yger]

Maybe I am misreading the graph, but to me this looks like a clear and nice unit here, and not sure there is any bug. Or am I missing something?

The fact that the y positions seems "noisy" does not mean that spikes are far apart. I guess the problem is more than estimating the positions on individual waveforms can be noisy, as function of the activity. The amplitudes over time is very stable, so I don't think this is a problem. You should check how do you estimate the position (monopolar triangulation, center of mass or grid convolution), and maybe try different methods. But the waveforms are all alike, so clearly, the spikes are not thousands of micron apart!

[aidapiccato]

You're right about the unit seeming otherwise to be a good one. I asked because I hadn't seen such a wide distribution of locations with other spike sorter outputs. I'm using center of mass for computing spike locations - I'll try monopolar_triangulation and see if that helps.

[samuelgarcia]

Hi Aida.
Thank you for this early adoption of lupin.
Feedback are really welcome.
I would have anwser the same as Pierre.

[alejoe91]

Hi @aidapiccato!
Another possibility is collisions! LUPIN might be correctly detecting colliding spikes, but in that case the "raw" waveform is compromised, so location estimation will give weird results (so woudl spike amplitudes, and you do have out of distributions points there too)