Using TrackObjects in Cellprofiler on multi-channel fluorescence movies

Usage & Issues
,
#1

Hi all,

I’m pretty new to Cellprofiler but really like the way it works. I recently worked on the image analysis of multi-channel fluorescence movies and thought this might be a great opportunity to get more familiar with Cellprofiler. During this ambition several questions arose that I couldn’t figure out on my own. In general:

  • I have .czi-files (Zeiss) with 4 channels and each 80 time frames.Each movie contains about 20 cells.
  • I used the ColorToGray module to split the channels and the TrackObjects module to track the cells over time.
    cp_pipeline_img-split
    cp_pipeline_img-split650×585 11.9 KB

Overall the segmentation and tracking works pretty fine and I’m really happy with the results.

However I do not exactly understand the parameters for the filtering of objects by lifetime since in the end I do have to filter the csv-results anyway on my own, right? At least even in the exported spreadsheet I see cells with less than my defined minimum lifetime. From what I see, I guess I have to filter for objects with “TrackObject_Lifetime_40” (40 in my case), right? Or are there any best practices to go ahead with the filtering of the data?

cp_pipeline_trackObj
cp_pipeline_trackObj1063×679 42.7 KB

Further I assume the per-image mean/median/standard deviation values for object measurements from the ExportToSpreadsheets module are therefore not correct or do they consider the correct objects with the defined lifetime?

In addition I tried two versions of Cellprofiler Analyst to look at the produced data, the Tracer version (https://cellprofiler.org/tracer) and the current standard build (https://github.com/CellProfiler/CellProfiler-Analyst). The tracer module itself works fine from what I’ve seen, but I was curious if I could use the CPA classification module to look at the tracked cells. Unfortunately I seem to struggle already at getting the objects fetched. The sample images are basically empty (not showing any cells). If I double click on a thumbnail, I get the following error message:

cpa_tracer_classifier_trouble1
cpa_tracer_classifier_trouble11379×1203 232 KB

With the current CPA build I get a different error, but it’s basically the same: empty thumbnails and an error on double-clicking them.

So I was wondering If I maybe made something wrong in loading or splitting the images in the original cellprofiler pipeline or if those kind of movies are not meant to be classified in CPA at all.

Any hints would be much appreciated and thanks for providing such a great tool!
Anna

#2

  1. Filtering tracks by fluorescence lifetime indeed happens at the end, and all that happens is that the track information is deleted; the objects still exist in the spreadsheet with all their non-tracking measurements (and still do contribute to per-image means, you’re exactly correct). If you want a database or spreadsheet that contains only tracked cells, you’ll need to create it yourself after, in Excel or a database management program or by using Python, R, etc.

  2. I’m not sure if CPA (at least the older builds with Tracer) can actually open CZI files in the way you’re intending; you may want to save out copies of the raw images as TIFFs or PNGs within CellProfiler (make sure to turn on saving the output location in SaveImages) if you want to do CPA classification.