Cell counting and Training in Analyst

cellprofiler

#1

Hi,
I’m new to CellProfiler. I know what I want to do but not how to do it. I have fluorescent images of cells I’d like to count and may classify if I could. The issue is that I also have a lot of cellular debris which is the same size, shape, and intensity as my positive cells so I don’t want false positives.

I believe steps are 1) Threshold on the intensity to highlight objects versus the background. I think I’ve gotten this far.
2) Keep a certain size range but exclude anything smaller or larger. I think there is a way to hover over objects in the image to find their size, this would help me figure out the right size range to keep. I haven’t figured out how to do this yet.
3) Set some parameters around shape. Things that are circular are probably what I want but things linear are not. Also things rectangular or triangular are not. Thing that look like a bowling pin or a snowman with a body and a head are. So this is where it gets tricky, how do I keep some non circular things but not others.

Also, if it could keep track of how many were circular vs. bowling pin shaped that would be awesome.

  1. If I could do the first three steps approximately well then I think I could use analyst to train the system which are the positives and which are the false positives. Or maybe I don’t even have to do the first three steps, maybe I could skip right to the training. But, I haven’t been able to do anything in analyst.

I have images I could post and a pipeline I’ve started, but I’m not seeing how to attach them.

Thanks in advance for your help.


#2

Hi,

There’s no way to hover over objects to get their size but what I tend to do for stuff like this use the DisplayDataOnImage module to display the sizes of the objects on the image (you will need to measure them first using MeasureObjectSizeShape). I then use the output to pan around the image to see the values for the objects.

It might take some time working out the various measurements that you get from MeasureObjectSize but hopefully you would find something that stratifies between what you want and what you don’t. If you do this you could use a FilterObjects module to separate them.

If you want to try and post a pipeline again I think that will help with getting an answer. You may also need to provide some sort of annotated image of what you count as the shapes you describe or what is positive/ negative because looking at those images I couldn’t work out what would belong to those categories.