Excluding broken channels

[NoahMoore]

I have a question that I've not found answered - apologies if it has already been asked. Basically, I am not sure how to deal with broken/floating channels in my silicon probe. One option is to keep them on and just remove clusters that are based on them, but I am assuming this would cause problems with the drift correction? Alternatively, I've tried hiding them from kilosort, but that then has implications for my later scripts which calculate how close cells are - for example, if two cells are two channels apart but the channel between them is broken, my scripts will assume they are one channel apart. So I wanted to ask if there is an option for kilosort to be aware that there is a channel there, but just disregard its data?

Thank you so much for your time.

[DanEgert]

I compared a few KS3 sorts with broken channels disabled or not (via the "connected" variable in channelmap) and got results that seem more realistic with them disabled. The sorts were definitely much faster. I think the geometry of the remaining channels in your channelmap should allow you to estimate the location of your cell even when channels are disabled.

My issue has been that once channels are disabled via the channelmap, the sort output refers to the filtered and whitened file generated, that only includes these channels. So for further processing it has to be kept around, which requires a lot of extra space. I don't suppose there's a way to label broken channels while having the sort output remain compatible with the original raw data?