How to set-up pipeline to distinuish between shapes?

Hi,

I am interested in using cell profiler to look at the difference between cells shapes in one image. Basically would like to quantify what is the % of shape 1 to % of shape 2. Is there a way to do this with cell profiler? If not, I could run two separate modules to get cell count of each shape in the sample image, how can I get cell profiler to count one cell shape and not another (distinct shapes) as both cell types will express the fluorescence? Any help would be welcomed, thanks!

Hi,

I think you’ll have to give us some examples to get the best help here. What you describe could range from a complicated, object- or model-based segmentation which we only do in our Worm Toolbox, to a simple measure a single shape parameter and find a cutoff value to determine your two classes. But without examples, I can’t guess what the feature might be. Take a look at the module MeasureObjectSizeShape (cellprofiler.org/CPmanual/Me … Shape.html) to get an idea of the shape features CP can measure.

Cheers,
David

P.S. I meant to say – Feel free to post a few image examples here, and maybe a stab at a pipeline, and we can evaluate it.

Thank you David for your reply! As I am kind of new to cell profiler, and usually used it in the past to quantify labeled cells only, I’ve never really played around with designing/modifying pipelines.

So, I will post an image and maybe you, or others more proficient, may be able to help guide me in the right direction. Basically, I am staining red blood cells, and looking at the % sickled vs. non-sickled. They are technically the same size, just misshaped.


I will take a look at measureobjectsizeshape module once I can get cell profiler to run on my new laptop:( It wont start the program, and states that my version of windows is unrecognized (Win 8). However, that is for another post, and I will have IT figure it out for now…

Thank you again for your help!

Hi,

I just created a short pipeline to start to classify cell shapes in your image. See attached pipeline.

Certainly the IdentifyPrimaryObjects settings might need to be adjusted, as well as the cutoff in ClassifyObjects. I chose Eccentricity as the feature to classify on, but you might find Compactness, FormFactor, or difference between the Feret istances might work out better for you. Test a little and choose one.

Side notes:

  • We suggest you not use JPG files because they are a lossy format. Rather, try to use PNG or TIFF.
  • You supplied a color image, but a grayscale one will be more efficient. Try and save directly into grayscale.

Hope that helps,
David
DL_classify_shapes.cppipe (9.08 KB)

Thank you so much David, your pipeline was very helpful along with your tips! I’ve been playing around with some settings to try and optimize. I really appreciate your help and advice!

Quick question, I’ve also been trying this pipeline on bright field images(minus the convert to gray module), will cell profiler work well with non tagged images. I’m finding some images spit out good numbers, but others have a lot of skipped cells that should have been included according to set parameters.
Thanks again for your efforts!

Hi,

Sorry for the late response, but bright field images are typically more difficult to analyze than fluorescent: viewtopic.php?f=14&t=806#p4488

Hope that helps,
David