Threshold determined

73 views
Skip to first unread message

Mo Hask

unread,
Nov 30, 2021, 3:51:54 AM11/30/21
to SpyKING CIRCUS
Dear Pierre,

I have a question about the thresholding step in SC. After running the algorithm with different thresholds (from 4 till 9) I find that SC always detects a relatively high number of spikes when compared to a normal peak detection function. The number of detected spikes by a simple peak detection function resemble more the ground truth data than spikes detected by SC which can be more than triple the amount of ground truth spikes. Is this on purpose in order to be sure that all spikes are detected or I just  have to find a threshold that best fits my data? Also I've seen that I should contact you if I want to use the BEER estimate. I am aware that the BEER should not be used in the case of bursting neurons or overlapping spikes.

Best,
Mohamed Alhaskir 

Pierre Yger

unread,
Dec 1, 2021, 12:13:11 PM12/1/21
to Mo Hask, SpyKING CIRCUS
Dear Mohamed

SpyKING CIRCUS uses a scipy function to detect peaks, so I'm surprised when you say that the number of detected peaks is higher compared to a "normal" peak detection function. You mean per channel, or by using the thresholding mode and thus multiunit activity? Because if you do use this mode, then of course you are likely to overestimate the number of spikes, counting some several times : if your probe layout is dense, you need to spike sort and process channels simultaneously to avoid this. 

We can skype if you want, since using the BEER will require some explanations. But to be honest, it might also be faster to generate artificial recording with MEArec and test performances with spikeinterface (these tools were not existing at the time of our paper). Depend what you want to quantify exactly

Best

Pierre


--
You received this message because you are subscribed to the Google Groups "SpyKING CIRCUS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spyking-circus-u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/spyking-circus-users/3976815a-6a42-40ef-997f-f7c41e9a3e62n%40googlegroups.com.

Mo Hask

unread,
Dec 7, 2021, 9:32:46 AM12/7/21
to SpyKING CIRCUS
Dear  Pierre,

Thank you very much for the fast response and sorry for the delayed answer. It would be great if we can meet for a skype call in order to  discuss more the beer estimate if you have the time this week or the next one.

Best regards,
Mohamed Alhaskir

Reply all
Reply to author
Forward
0 new messages