search

denisecailab / minian

miniscope analysis pipeline with interactive visualizations
GNU General Public License v3.0
91 stars 35 forks source link

Questions about the running results #237

Open lance11320 opened 1 year ago

lance11320 commented 1 year ago

Hi, thanks for your great work first! It is quite user friendly! I had a few questions when using it:

  1. It seems that in saving period, the matrix C saved is a smooth fitted one. And I cannot find where I can save the original signal trace. Maybe I missed it in docs.
  2. I tried it in some recording sessions, and find that the extracted footprints have overlapping. This seems to be abnormal and I don't know if it is due to the parameter settings. I used the default settings. (when computing, there will be some warning saying that "divided by zero error, solving sub optimally", is this related to the abnormal results?)
Nate-the-Neuroscientist commented 1 year ago

CNMF is the neuron-extraction backbone of minian. Unlike it's predecessors (such as PCA/ICA), CNMF allows for neurons to overlap in 2D space, which they often do. So over-lapping neurons is expected from the minian pipeline.

Similarly, I think that CNMF may depend on the smoothened calcium trace in order to extract the neural activity (C matrix), which leads me to think that it may not be possible to obtain the signal traces with the noise, but I am not sure about this...