ciatah icon indicating copy to clipboard operation
ciatah copied to clipboard

parameters in PCA-ICA analysis

Open NataliaFavila opened this issue 1 year ago • 6 comments

Dear Biafra,

we are using Ciatah to analyse our Ca imaging data, but we are struggling finding good parameter values for the PCA-ICA cell extraction. Can you tell me what the mu and term_tol paramters are? or point me towards any documentation or literature where they are explained?

Thank you! Natalia

NataliaFavila avatar Jul 25 '22 13:07 NataliaFavila

Hi Natalia,

Some explanations below. I'll add some of this information to the docs (https://bahanonu.github.io/ciatah/pipeline_detailed_signal_extraction/).

As a quick question: what type of data are you analyzing and do you have a snippet of a representative movie you are trying to analyze or can share a frame? That'll also help determine whether anything can be done on the pre-processing end to improve PCA-ICA runs or if the data will work well with PCA-ICA.

Mukamel, 2009 (mu)

The original Mukamel, 2009 (https://doi.org/10.1016/j.neuron.2009.08.009) paper describing PCA-ICA gives an explanation of mu:

image

If you look at Fig. S3, it also provides some information on the effects that varying mu 0 to 1 have on cell extraction quality (we've often found lower values, e.g. 0.1, work well):

image

Ahanonu, 2022 (mu)

We also describe it in our recent calcium imaging experiments and analysis book chapter, see section 3.15 Extraction of Neuron Shapes, Locations, and Activity Traces: https://link.springer.com/protocol/10.1007/978-1-0716-2039-7_13#Sec19.

image

term_tol

The term_tol parameter is the termination tolerance, e.g. when min difference between ICA iterations is below this value, the algorithm will exit (if it hasn't already reached max_iter).

cheers! -Biafra

bahanonu avatar Jul 25 '22 14:07 bahanonu

Hi Biafra,

thank you very much for all the info, it is very helpful. We are analysing 1p data of interneurons (obtained with GCaMP6f), so the signal-to-noise ratio is not super good. Any recommendations will be much appreciated.

thanks!! Natalia frame_example

NataliaFavila avatar Jul 26 '22 09:07 NataliaFavila

Glad to hear!

Would you be able to provide a link (e.g. Dropbox/Box) to one of the movies? I can then take a look and get back with pre-processing/cell extraction suggestions.

bahanonu avatar Jul 26 '22 14:07 bahanonu

Hi Biafra,

here is a link to one of the movies[movie_exmaple] (https://drive.google.com/file/d/1ROwykUCIMlNQtzyCrNj_g2ja279q2AVI/view?usp=sharing)

Our videos are already spatially filtered and spatially downsample 4x (so the movies are 270 x 270). Thank you very much for the help! Natalia

NataliaFavila avatar Jul 27 '22 09:07 NataliaFavila

Thanks, will take a look. Could you send a movie without spatial filtering? Will make it easier to test pre-processing on my end (e.g. spatial filtering for motion correction can be different than final spatial filter applied to the movie).

bahanonu avatar Jul 27 '22 13:07 bahanonu

Hi Biafra, unfortunately we don't have the original inscopix videos :/ Thank you so much for your help!

NataliaFavila avatar Jul 28 '22 11:07 NataliaFavila