Positive crossing detection

[bruno-f-cruz]

Hi!
First of all thanks for the really useful piece of software you have built.
We are a bit confused about the interaction of kilosort with positive spikes. In the wiki it reads:
"First we determine which channels contain significant numbers of negative threshold crossings ". However when we are looking in phy, very often we encounter waveforms like the following:

Interestingly, however, these cells tend to get hard cut from the noise very frequently. Hence my question is: does the amplitude view calculate the overall amplitude of the waveform (be it positive or negative) OR does it only take into consideration the negative deflection component of the waveform?

[marius10p]

The step you point out to find good channels is independent of how the rest of Kilosort2 works. The amplitudes are the positive scalar values that multiply the templates to best match the raw data spike. The reason your axonal templates get cut off is that they have very small spatial extent, and they therefore get lost in-between the sites as the probe drifts. You can even see them emerging from the noise as they come into view when the probe drifts. There is nothing anyone can do here I'm afraid to recover the data in-between channels. You can try reducing the amplitude even further (it's already very low at 4) but I doubt you'll be better able to separate them from the background.

[marius10p]

Closing for inactivity.