I have recently begun analyzing electrophysiology data and I want to make sure that I am recording on true events above the noise of the recording.
Is there a way to determine a good threshold setting. Some events are clearly true responses, but some are smaller and may still be events. How does one determine this?
Is there a 'Rule of Thumb'?
Dear Cody,
I suggest you to see links and attached files on topic.
TRP Channels - Page 287 - Résultats Google Recherche de Livres https://books.google.dz/books?isbn=1439818614
Age related changes in tonic activation of presynaptic vs. extrasynaptic ...
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5235322/
Novel Use of Matched Filtering for Synaptic Event Detection ... - PLOS
https://journals.plos.org/plosone/article?id=10.1371/journal...
Integration of inhibitory and excitatory effects of α7 nicotinic ...
https://www.sciencedirect.com/science/article/.../S00283908163006...
Best regards
Hi Cody,
Welcome to the world of minis. In 90% of neurons, minis will be distributed with such an amplitude so that the smaller ones are too small to be detected in the noise. And there is nothing you can do about this.
Typically, despite their being better approaches [for example that manuscript from Paninski cited above] that give some statistical rigour to the detection of minis, most people just "do their best". Typically, especially if you're most interested in the frequency of minis you set your threshold so that you detect everything that could be a mini, and a bunch of random junk, then you go through manually and curate the results. Of course, when I say "threshold" here, I am talking about the output of some template matching algorithm, rather than an amplitude threshold, as you will not detect many minis if you use an amplitude threshold. When it comes to what to keep and what to throw away, it is difficult. You just have to use your best judgement.
Dear Cody,
I second William Connelly. With minianalysis and most softwares to detect events, a lot of the smaller events will not be detected by the software and it appears to be an unavoidable problem. You could set the threshold low enough to detect the events and then revisit to check the accuracy of detection.
Hi Cody,
There really is no correct threshold as PSC amplitude distribution is not separated from the amplitude distribution of noise fluctuations. Therefore, the reported studies in the literature miss 90% of all PSCs on average. You can find the discussion of the issue in Martynas Dervinis, Guy Major
bioRxiv 2022.03.20.485046; doi: https://doi.org/10.1101/2022.03.20.485046