The way to discriminate (reject) pseudo (false) spike?

[MBLH]

The way to discriminate (reject) pseudo (false) spike?

[ferchaure]

I'm not an expert in tetrode signals... but I will try to help

case 1: Do you check using only a negative threshold? Using both types doesn't help in you case. Worst case scenario, you can make two time, one with each type of threshold. Why don't you search for cluster in another temperatures? maybe the ugly clusters are better at other temperatures.

case 2: But... cluster 2 could be a neuron... maybe isn't clear in all the electrodes but you shouldn't reject them because one electrode is far away of the neuron.
The ISI and the spikes of cluster 3 are quite odd... maybe noise. You can try to find the pure noise cluster with manual or choosing another temperature. If you make a cluster with all that noise, it won't bother any more.

case 3: Again, the threshold aren't helping. Maybe, if you change par.stdmax and remove the noise, the plots will be more easy to understand.

[seble2016]

Hi Ferchaure,
I'm leaving my comment here as my question follows this thread already.
Could you explain to me how one decides on temperatures? Is it arbitrary?

[ferchaure]

A few rules of thumb about temperature selection:

• If you see a big increase in the size of the clusters when you increase the temperature, check the temperature after the increase. Because new clusters appeared and that could be neurons.
• Check the lower temperature for clusters with more than a few dozens of spikes. Noise or sparse neurons with a characterised waveform can be there.
• Remember that each temperature is a possible clustering of the data. If you found a nice cluster fix it always before change the temperature.
• Sometimes you will find clusters in different temperatures with waveform too similar, merge them if you think that the difference is to close to the noise in the signal.