Cross-session cell alignment, signalSorter, and misc updates

bahanonu / ciatah

CIAtah (pronounced cheetah): a software package for calcium imaging analysis of one- and two-photon imaging datasets. Documentation: https://git.io/ciatah_docs. Formerly known as calciumImagingAnalysis (ciapkg).

https://git.io/ciatah_docs

MIT License

80 stars 20 forks source link

Cross-session cell alignment, signalSorter, and misc updates #19

Closed bahanonu closed 5 years ago

bahanonu commented 5 years ago

Cross-session now supports user sessions that don't have names recognized by getFileInfo.
signalSorter supports input of secondary trace for use with trace and ROI viewer (r in the GUI).
moveFilesToFolders added to the repo. Used with modelDownsampleRawMovies.

bahanonu commented 5 years ago

@omidm7

See this update, tested on the data you sent me and worked fine.

omidm7 commented 5 years ago

Thanks Biafra. I'm trying to run preprocessing with the new function updates but keep running into this error:

It results in Fiji closing down, and no cropped output files.

bahanonu commented 5 years ago

@omidm7, refresh with most recent commit; tested on example data and don't run into that problem.

omidm7 commented 5 years ago

Awesome, I made it all the way through the pipeline. I even made it mostly through cross-session alignment, but hit this error near the end:

bahanonu commented 5 years ago

What settings are you inputting to computeMatchObjBtwnTrials (e.g. see below)? And if I recall correctly, you are using 2017b on OSX?

bahanonu commented 5 years ago

Also, if you run on the same dataset of 3 sessions you sent me previously, do you run into any issues?

omidm7 commented 5 years ago

I am using the default settings (1 round, 5 px, 0.4 threshold, blank, 2). I'm actually trying to align between two of the three data sets I sent you. (See folders below.)

bahanonu commented 5 years ago

@omidm7

Found the issue, appears some CNMF-E images during registration end up containing all zeros, causing a divide by zero issue (didn't encounter on Windows, but testing 2017a/2018b there, so might be specific to a low-level Matlab function in OSX 2017b). Fixed.

Check again with most recent pull request, ran it on that dataset using 2017b on OSX 10.10 and worked fine.

omidm7 commented 5 years ago

Do I need to run CNMF-e again for the fix to work? I tried cross-session alignment with the latest pull request and still got the same error.

bahanonu commented 5 years ago

Odd, and you are running this on OSX Matlab 2017b with the same data that you sent me via Box? I could do a quick remote desktop session to check if there is something else if that'd be easier.

omidm7 commented 5 years ago

Yup! I just downloaded teamviewer. Lmk when you want to hop on!

bahanonu commented 5 years ago

Sounds good! Busy today, I'll shoot you an email when I'm free.

bahanonu commented 5 years ago

@omidm7

Checked code on your computer via Teamviewer and runs fine through computeMatchObjBtwnTrials and viewMatchObjBtwnSessions. I'll close this pull request pending no other issues on your side.

omidm7 commented 5 years ago

Thank you. Everything works fine on my end now.

omidm7 commented 5 years ago

Hey Biafra, question. In viewMatchObjbtwSessions there is a Global ID followed by 5 numbers. I assume those are the cell numbers in their respective sessions. (See below.) Is this information available as an output somewhere? I tried modelSaveMatchObjBtwnTrials to see if that would create an output file, but got the following error:

bahanonu commented 5 years ago

@omidm7

Can find the matrix with matching session IDs for the same cell across sessions at obj.globalIDs.m01. See below for description.

[M N] matrix with M = number of global IDs and N = each session. Each m,n pair specifies the index of that global obj m in the data of session n. If .globalIDs(m,n)==0, means no match was found.

modelSaveMatchObjBtwnTrials isn't supported yet (making it so users can more easily export the cross-session information), will add it in the next pull request.

omidm7 commented 5 years ago

Hey Biafra, all is working well. Thanks. One suggestion: A 'go to global ID' feature in viewMatchObjectsBetweenTrials would be super useful! Similar to the 'go to signal' feature you have in computeManualSortSignal. Right now I am using the left/right cursor to move through signals, but with 475 global IDs, it gets tough. :)

Omid

omidm7 commented 5 years ago

Also, there seems to be a disconnect between the global ID index given by obj.globalIDs.m01 vs the global ID index given by viewMatchObjectsBetweenTrials. For example, in the variables section below, global ID 3 should be assigned to cells in sessions 1, 3 and 4. But in the figure, the global ID list assigns global ID 3 to all five sessions.

bahanonu commented 5 years ago

@omidm7

Great to hear. Yep! Will update that interface at some point to make more user friendly, been meaning to add some features/more information.

The interface sorts globalIDs by the maximum number of objects aligned, so it won't necessarily match that in obj.globalIDs.m01. I can change this as this may be confusing to some users