FEATURE: Correcting for misaligned spikes within units

[dimokaramanlis]

Feature you'd like to see:

Hi Marius and KS team,

we have been testing KS4 with NP1.0 probes and we are extremely happy with the results so far.

However, sometimes units seem to contain spike times that are misaligned in respect to their corresponding waveforms, typically forming bimodal distributions in feature spaces. See example below:

Default parameters were used. I guess this bug arises during the merging step.

While for most analyses these small differences in timing don't matter, it can be helpful to have aligned waveforms for calculating average spike shapes for feature extraction (e.g., peak-to-trough times).

Best, Dimos

Additional Context

No response

[hemant22]

I also see this misalignment in many of my units, giving two well separated clusters

[JuanPimientoCaicedo]

Hi, I would like to add the same thing. This is happening with Kilosort 4, I would say in my case it happens around 5% of the times. Here is an example:

[jacobpennington]

Thanks for the extra examples! We're working on a fix.