To distinguish between useful clusters and noisy clusters in Klusta

[Shavika Rastogi]

Hello everyone !

Since I am new to spike sorting, I want to have better understanding of how it works.

I am doing spike sorting using klusta on the extracellular recordings recorded from different cortical depths in rat brain. Phy gui shows different clusters obtained as a result of spike sorting as well as number of spikes in each cluster. However, there is always one cluster in which number of spikes are very very large as compared to rest of the clusters.

Should this cluster be considered as a noise clustered and thus eliminated from our calculation of firing rates ?

I have attached the screenshot of the results obtained using klusta in Phy gui. I can be seen that number of spikes are very,very large for cluster 2 as compared to other clusters.

Hoping for an early response.

Regards

Shavika Rastogi

Visting Research Student

Biocomputation Research Group

University of Hertfordshire, UK

[Michael Okun]

The waveforms in cluster 2 look like noise, and nothing like actual spikes ( just one point in time appears to cross the threshold).  If this is a result of having the spike threshold too low, it's worth increasing it.