I am quite new to CellProfiler. My aim is to identify all desmin-positive myoblasts (cytoplasmic stain) in a cell population stained with HOECHST. For establishing a pipeline I used the published example percent positive as a starting point. However, I end up with two problems and I am not sure how to solve it.
(1) The cells often attach to or even merge into each other. Do you have some suggestions to solve that better than I did?
(2) The cells are myoblasts and often differ in size a lot since some of the myoblasts already start to fuse into multinucleated cells. Thus, in most of the cases, the number of nuclei is higher than the number of myoblasts (cells) identified even if all cells are in fact desmin positive. This would give me wrong percentages but also false negative results. However, size is not a good criteria because mononucleated myoblasts can also vary in size a lot and can be even bigger in size sometimes (it is just a primary cell line ). Do you have any idea how to handle this?
I added two pipelines with different classifications and example figures. The example figures contains some nuclei which are not surrounded by desmin stain, but most of them are.
Can anybody have a look and give me some advice?
Thanks a lot!!!
Regards from the muscle site
Desmin10x_19052015_2.cppipe (20.6 KB)
Desmin10x_19052015.cppipe (20.6 KB)
I’m attaching a pipeline that should be a decent start. You basically had the right idea, but I added a ReassignObjectNumbers module to attempt to unify nuclei that are actually from a multi-nucleated myoblast. Then MaskObjects is used to check which nuclei overlap by more than a given percentage with a desmin+ region, and filter on that basis.
I emphasize “attempt” in ReassignObjectNumbers since I don’t know the exact criteria to deem two neighboring nuclei as members of the same cell. That is:
- Do they need to be touching?
- Are they seaparted by some maximum distance?
- Can a cell be multi-nucleated but have one portion desmin+ and the other desmin-?
In any case, I tried to specify a maximum distance in ReassignObjectNumbers and then use the desmin image as the guiding image in order to unify nuclei that are within a certain distance and have a sufficiently high desmin intensity along a line connecting the centroids nuclei under consideration. You may need to play around with this further.
assay.cppipe (11.8 KB)
thanks a lot for your reply and sorry for the delay in my answer.
Regarding your questions for the criteria to deem two neighboring nuclei as members of the same cell: They do not need to be touching - but sometimes they do and sometimes not within the same cell. In myoblasts the nuclei do not have to be in the middle of the cell or do not have to be completely surrounded by the cytosol. Thus I do not think that there is a reasonble maximum distance. However, a multinucleated cell is always completely desmin positive. Desmin-negative cells are never multinucleated cells! Anyway, most of the desmin-positive cells have only one nucleus.
However, then I read your answer I was wondering whether it would be possible to analyse the pictures the other way around - meaning that we do not count desmin positive cell but count all nuclei which are not surrounded by any desmin stain. The result would give a percentage of nuclei surrounded or not surrounded by desmin. Would that be possible? And if so, how can I implement that into the pipeline?
Thanks al lot!
This is actually a straightforward issue: To count desmin- cells, in the FilterObjects module, just reverse the selection criteria. That is, since desmin+ nuclei have a child count > 0, desmin- nuclei have a child count equal to 0.