Dealing with bleaching

[zhouyi0812]

Hello!
My movies have different degree of bleaching and I check there are different ways of dealing with it.
I am wondering how do you usually deal with it? I saw there is a debleach(self) function on the core function of caiman, if that is enough?
Thank you!

[EricThomson]

Once you extract a trace, applying detrend_df_f() will remove artifacts on long time scales like bleaching. That debleach() function is an interesting discovery: I have never seen it before. 😮 The movie object used to be a catch-all for many different methods, and this seems to be a vestige of those days. I'm not sure it should even still be in the code base 😄

[zhouyi0812]

Thank you!
Just wonder if the right order would be: Getting denoised trace: estimates.C --> remove bleaching: detrend_df_f --> normalize trace to get df/f by z score

[EricThomson]

That would be a reasonable strategy.

[zhouyi0812]

Hello!
May I ask a follow up question?
How to know if detrend_df_f is using the denoised data?
The demo is directly use cnm.estimates.detrend_df_f(quantileMin=8,
frames_window=540,
use_residuals=False),
I was trying to cnm.estimates.detrend_df_f(denoised_trace[quantileMin=8,
frames_window=540,
use_residuals=False]) but this is not necessary.
Thank you!

[EricThomson]

The algorithm will automatically use denoised data if use_residuals is False (for 1p data it will just detrend, not normalize).

[zhouyi0812]

Hello! I wonder after I apply detrend_df_f on data, the output is caiman.source_extraction.cnmf.cnmf.CNMF.
And if I want to get traces, can I go with this?
Thank you!!
good_inds = cnmfe_model.estimates.idx_components
denoised_traces = cnmfe_model.estimates.C[good_inds,:]