Need Help with Data export and reassigning filtered objects

Hello,

since it has been a while since I posted on this forum, I open a new thread. The old one can be found here:

viewtopic.php?f=14&t=1102&p=4472&hilit=roeper#p4472

Again, thanks a lot for all that input. I worked a lot on my pipeline and also had to make some changes to my actual experimental setup. Since I had big issues with Antibody-staining as well es Autoflourescence on human vascular smooth muscle cells, I switched to VSMC from GFP-mice, which work well for what I am doing.

I build a more advanced pipeline, already generated about 600 Gigabyte of Image Data and I am almost ready to analyze them :smiley: I just got a little issue with exporting statistics. Since I cannot fully analyze my pictures automatically, I use the Cellprofiler to filter potential events, sort out those images and export a modified version of the picture that I can easily and fast double check. However, I will need the total number of “VSMCNuclei” and potentially “MacNuclei” as well, so I can do the statistics later and determine the percentage of positive events compareds to the total count of VSMC. Right now, I have a spreadsheet which gives this data by counting the rows of entries. However, I will have hundreds of thousands of cells and I wondered if there is a simple solution that can give me the total number of the desired cells after each run (which consists of about 1200 images).

I got another thing that would still be able to improve my pipeline: Is there a way to make those objects (VSMCNucleiUnfiltered) become “MacNuclei” that are filtered in step 7? (All VSMCs that are too bright are most likely to be big Macrophages, so I would rather treat them as those instead of deleting them from the picture. It will not affect the total count a lot, but if one can do that, it’d be great.

I included a zip file with 5 test images as well as the pipeline and would appreciate any help.

Thx and best regards,

Matthias Roeper
Roeper - CP - 20110802.zip (9.11 MB)

The count of VSMCNuclei and MacNuclei will be included as per-image data if you use ExportToSpreadsheet or ExportToDatabase. The column names will be along the lines of “Image_Count_VSMCNuclei” and “Image_Count_MacNuclei”.

To clarify: Are you asking whether the nuclei that < 700 in area be combined with the nuclei that are < 700 in area and > 0.2 in intensity? If so, it’s doable, but but a bit complicated:

  • Use ConvertObjectsToImage to change the requiste objects back to a pair of grayscale images

  • ImageMath to add the grayscale images together (and combine the objects)

  • Use IdentifyPrimaryObjects to re-identify the nuclei from the summed image.

If this is what you are doing, I wonder whether you might be better off using the machine-learning tool in CellProfiler Analyst (CPA). In that case, you would collect as many measurements as you can on the nuclei, point CPA to the measurements and sort some example VSMCs and macrophages into bins so the computer can figure out the difference, then score all the images once the classifier is accurate enough. The output at the end is a per-image (or per-well, if you include well metadata collected in LoadImages) table of total cell counts plus VSMCs and macrophage cell counts. Getting CPA configured can be a bit tricky but it’s a powerful tool and we have a CPA forum as a resource as well.

Regards,
-Mark