search

MouseLand / Kilosort

Fast spike sorting with drift correction for up to a thousand channels
https://kilosort.readthedocs.io/en/latest/
GNU General Public License v3.0
456 stars 238 forks source link

Value of amplitude #697

Closed Lorena-J closed 3 months ago

Lorena-J commented 3 months ago

Describe the issue:

Hello! I have been testing ks4 with a single shank probe and comparing the rating with ks 2.5 and 4 there is a big difference in the amplitude values and I don't know what are the units of this amplitude in ks4:

Captura de pantalla 2024-05-23 a las 12 00 46
jacobpennington commented 3 months ago

Hello,

Kilosort4 does not save amplitudes in exactly the same way that past versions did, so there will be differences. Please see my last reply in #632 for details. However, they should not be nearly as large as you've shown in that screenshot, that's many orders of magnitude larger than expected. Are you using the latest version of Kilosort4? There were updates to the amplitude calculations not too long ago.

Lorena-J commented 3 months ago

I am using the lastet version, but is there a problem with other parameters, or can I do the classification and calculate the amplitudes myself?

jacobpennington commented 3 months ago

I would assume there are issues with other parameters, but I can't tell without seeing some screenshots and/or printouts. To give you an idea, the scale of the 'amp' column (which is computed by Phy based on the templates that Kilosort learns) is around 100-500 for our sample Neuropixels 1 dataset. That scale is expected to be different for different datasets, but that big of a difference seems very unlikely.

Can you please post a screenshot of what the full GUI looks for you prior to sorting, including the "extra settings" window? I.e. something like this: image

Copy-pasting the full print output generated during sorting would also be helpful, if it's feasible to run it again.

Lorena-J commented 3 months ago

yes, sure, I'm doing the analysis with a single shank probe, it has 64 channels, I don't know why but channel '0' is not represented in the probe view, here are the screenshots:

Captura de pantalla 2024-05-27 a las 12 33 18 Captura de pantalla 2024-05-27 a las 13 27 20 Captura de pantalla 2024-05-27 a las 13 10 40 Captura de pantalla 2024-05-27 a las 13 26 34

thank youu!!!

jacobpennington commented 3 months ago

If a channel is not showing up in the probe view, then it's either not included in the probe file or the coordinates were entered incorrectly (i.e. possibly overlapping another contact so that you can't see it). That could cause some of the preprocessing variables to behave unexpectedly, so you would need to fix your probe file first.

jacobpennington commented 3 months ago

Closing this for now, if you still encounter problems after updating the probe file please let us know.