Skip to content

Latest commit

 

History

History
123 lines (103 loc) · 5.41 KB

workflows.md

File metadata and controls

123 lines (103 loc) · 5.41 KB
Raw

vrAnalysis Documentation: workflows

This is documentation for standard workflows using the vrAnalysis package. The main goal of this file is to remind me what I've written so I don't forget. There may be room for explaining what is going on, but mostly it's just code blocks indicating standard usage.

Red Cell Quality Control Workflow

The red cell QC control workflow is pretty simple. It requires the following import statements and the database object:

from vrAnalysis import session
from vrAnalysis import database
from vrAnalysis.redgui import redCellGUI as rgui
vrdb = database.vrDatabase()

Red cell quality control uses four features computed automatically during registration, along with additional manual curation. There is a nice GUI that you can use to choose thresholds for each feature and to manually toggle whether a cell is red or not in addition to the feature criterion. The full documentation page for the GUI is here.

There are two general workflows for curation of red cell assignment. The first is to go through each session and do the curation independently of all other sessions. I refer to this as the "Primary Worklow", because it's pretty efficient and you always have to start with this.

The second method is to do the curation for one session and then apply it to other sessions automatically, (usually to sessions from the same mouse). This is obviously efficient, because it only requires direct curation from one session, but I recommend at least checking all sessions, especially because you can do a little extra manual curation after choosing the feature thresholds.

Primary Workflow

  1. Identify which sessions need red cell QC The database can create an iterable list of sessions that require red cell quality control, along with any other filters. I usually start by iterating through this list and printing the session names. Here, I'm filtering to only include sessions from one mouse, to focus curation on a single brain.
for ses in vrdb.iterSessionNeedRedCellQC(mouseName='ATL027'):
    print(ses.sessionPrint())
  1. Open up a red cell GUI viewer for one of the sessions Then, choose a session to curate, and open up the red cell GUI viewer. Note that I've only tested this in Jupyter lab. The following code block will open up a GUI based on PyQt5 and Napari.
mouseName = 'ATL027'
dateString = '2023-08-01'
sessionid = '701'

# Load registered vrExperiment
vrexp = session.vrExperiment(mouseName, dateString, sessionid)
redCell = session.redCellProcessing(vrexp)
redSelection = rgui.redSelectionGUI(redCell)

  1. Do curation in the GUI, save work and update database This all happens in the GUI. For information on how to use the GUI (it's a little tricky to use but powerful for my needs), go over to the redCellGUI documentation page.

  2. Check your work! You can also print a list of sessions that have finished quality control. I like doing this to sanity check that I curated successfully (where successfully means that I pressed the save and the update database button). Note that this only checks if you updated the database, saving the results is a different part of the GUI. (But of course you can always reopen the redCellGUI for the same session and make sure it looks good).

for ses in vrdb.iterSessions(mouseName='ATL027', redCellQC=True):
    print(ses.sessionPrint())

Secondary Workflow

  1. Perform the primary workflow for at least one session, then save. Once you do this, the threshold values for each feature will be stored in the one-data for the session you loaded.

  2. Apply the features to other sessions. Now, you can apply those threshold values you selected to other sessions. First, create a redCellProcessing object for the session you already curated. In the example, I curated session 'ATL027/2023-08-01/701'. Then, go through all sessions you want to copy the thresholds to with an iterator of sessions that need red cell QC (usually from the same mouse). Make a redCellProcessing object for each session, then use the updateFromSession method to copy threshold from one session to another. Note that I'm using autoload=False because redCellProcessing objects load data by default, which takes a few seconds, and we only need the class methods for this purpose.

copyCriterionFrom = session.redCellProcessing('ATL027','2023-08-01','701', autoload=False)
for ses in vrdb.iterSessionNeedRedCellQC(mouseName='ATL027'):
    redCell = session.redCellProcessing(ses, autoload=False)
    redCell.updateFromSession(copyCriterionFrom)

Additional Features

Sometimes you'll want to compare the cutoffs you chose for sessions in the same (or different) mice. To do so, use this block. Make a session iterable to go through a target list of sessions, generally with a filter on the <mouseName> and only looking at sessions where you've done red cell QC. It'll print a pandas dataframe showing the min/max cutoffs for each feature value used in the red cell detection. You may want to increase the display width of pandas before doing this in whatever notebook you're working in.

import pandas as pd
pd.options.display.width = 1000
rgui.compareFeatureCutoffs(*vrdb.iterSessions(mouseName='ATL027', redCellQC=True), roundValue=3)

Note that if you use the secondary workflow, then the feature cutoffs should be the same for every session. It's a good way to check your work!