vessels.cpproj (418.8 KB)
Hi all,I’m a new hand for Cellprofiler,I have stucked in the IdentifyPrimaryObject module for a few days,just as the picture I have uploaded,I want to Identify and quantitative blood vessel morphology(the brown part),but my design doesn’t .I have been studying related experience in cellprofiler help,but I really can’t make my project better.Looking forward to getting some help from you.
What specifically about the vessels do you want to quantify, and what specifically is not working the way you want it to? Thanks!
Please dowload this vessel_DAB.zip (4.8 MB) file that contains a project that can process your images and output some results.
The two conditions are so different that you could use a simple Area measurement (you should consider normalizing it using area of tissue or area of nuclei )
Some concerns regarding your dataset :
the nuclei are smaller in the bottom image.
Is that (A) a cell feature or (B) the images were taken with different magnifications ?
If it’s B you should consider taking new images using the same condition if you want to compare them.
regarding “blood vessel morphology” . One “classical” way is to skeletonize a binary image of the vessel and look at number if junctions , maximal length … An issue here is that your staining have very irregular shapes, holes and tissue looks damaged in some part. This result in artifactual increase of junctions number in the skeleton (look at the output images at end of the project)
Thanks and sorry for didn’t check my email in time.I’d made a tube,but I’m not sure it’s the right thing.Yes,the brown part in the picture I’ve uploaded is I want,the vessels.May I hope some suggestions from you?
Thanks for your time again!Best wishes!
Thanks for your sweet tips!I’ve downloaded your suggested file,but I get some troubles.Could you take some time check the tube I have uploaded?Awaiting for your reply!
Best wishes for you!OK.cpproj (422.3 KB)
I find your approach ( dividing one color by another) to highlight the vessels very nice
it could give varying results because of brown varying darkness (DAB is a non-linera scaterer of light and can be randomly more or less dark)
and since you have color deconvolution module available it would be easiser to convince potential reviewers .
You can also blur a little your resulting image to avoid the division into small pieces.