I am working on the segmentation of datasets from mammalian cells that form tight colonies and I am still relatively new to CellProfiler.
My dataset are timelapses of 160 timepoints (5 min interval) of 2 channels, a SiR-Hoechst nuclear stain and my protein of interest with GFP, that localizes to the adherence junctions, cytoplasm and nucleus (if stimulated).
I attached an example of 20 images (16-bit) from both channels, to illustrate what it looks like.
dataset 20 timesteps.zip (5.9 MB)
So far I have been working with this pipeline
pipeline_for image forum.cppipe (31.7 KB)
I have two major problems still, and I would be very grateful for any advice!
The segmentation works OK, but occasionally the cells are so tightly packed that it recognizes multiple nuclei as only 1. The main issue with this is that the merging nuclei in the subsequent tracking all get the same number when they merge, and keep this same number after they split again.
For example in these 20 frames cell 34 and 35 (lower left cells in the big colony) merge at T=8 & T=9 and in T=10 they are two objects again, but they will remain to be called cell 35 from then on.
Montage nuclei T=7-10.tif (768.2 KB)
- Is there a way to still improve my segmentation settings? I tried many options already, but maybe I am missing something.
- Is there a way to take into account the timelapse to do a guided segmentation (i.e. If in T=1 there are 25 objects, it tries to find these 25 objects in T=2 as well?). Of course I have some appearing, disappearing and dividing cells, but it seems to me that the integration of timepoints (if at all possible) could aid the segmentation hugely.
- Would it be useful to use the function SplitOrMergeObjects to somehow get better management of the tracking after a clumping incident? Would anyone have recommendations?
- I also tried to start form the other channel (With the adherence junction) since this visually indicates the cells very well, but this did not work at all so far.
My second problem is that the tracking of the objects is not linked. The tracking for the nucleus (except for the above described issue) works quite alright. However, in my current pipeline the tracking of the cytoplasm and membrane is a separate module. I don’t think this tracking is very effective and after time the cytoplasm/membrane don’t have the same track label as the nucleus they belong to anymore. Here is an example at T=1 (where everything still corresponds) and T=10 of the example data. For example in T=10 many cells are called 9, whereas there is only 1 nucleus called 9 (top right)
As the membrane/cytoplasm are tertiary objects (that come from a secondary object whole cell) they should always correspond to a nucleus after segmentation, but I don’t know how I can use this link to inform that the track label should just always correspond to the one from the nucleus.
- Is it possible to just track the nucleus and then use the link that was made in the segmentation to get tracking of all my compartments so that at least these track labels always correspond between the different compartments?
Thank you so much,