Selecting specific (non-random) spikes for waveform calculations

[Sietse-k]

Discussed in #3560

^{Originally posted by Sietse-k December 2, 2024}
Hi,

I want to compare the waveforms of units in different periods of the recording. Basically, do the waveforms of the same unit look similar during optogenic stimulation periods vs. outside these windows.

Instead of selecting spikes "randomly" in the recording, is there a way to specify which spikes you want to select (based on time intervals)? Or would I always need to slice the recording first into a smaller "sub recording" and then select all spikes within this sub recording, and then use this to compute the waveforms?

analyzer.compute(
    "random_spikes",
    method="uniform",
    max_spikes_per_unit=500,
)
analyzer.compute("waveforms", ms_before=1.0, ms_after=2.0)
analyzer.compute("templates", operators=["average", "median", "std"])

Any help would be much appreciated,

Sietse

[zm711]

Hey @Sietse-k,

currently the options are method="uniform" or method='all'. The plan was to extend these methods so users could choose other methods of selecting spikes but we haven't finished it yet. Maybe @samuelgarcia or @alejoe91 have a decent strategy you could try in the mean time, but this is on our todo list (just not sure how high up it is for now).

[alejoe91]

@Sietse-k there are 2 strategies:

1. Create multiple sorting objects filtering the spikes inside/outside of your intervales. You can do it easily with the sorting.to_spike_vector() function! Then create multiple analyzers and use all random spikes method.
2. Create a single sorting object with to spike trains per unit. One woul be {unit_id}_inside, the second {unit_id}_outside. then you can create a single analyzer since templates and extensions are computed per unit.

Let me know if it's clear!!

[Sietse-k]

Hi @zm711 & @alejoe91. Thanks a lot for the quick replies, I will try both strategies!