In need of much help isolating plant cells. Yelp

Hi David or Mark or any other altruistic soul,

I am a complete newb here. I’ve been messing around with your software for the past couple days to no avail, and I could really use some help.

I am trying to identify and measure the area of each individual bundle sheath cell of a leaf cross section. They are the the circular objects arranged in a ring-like fashion around a darker core (leaf vein) in the attached picture. I’ve been using the ExampleSBS-IlluminationPIPE.mat pipeline as instructed in the your PDF titled “Extracting Rich Information from Images” to see if I could even get anything to materialize, but I’ve been really struggling.

I know it’s easier to take example pipelines and tweak them for our purposes, but at this point I’m not even sure I’m using the correct pipeline. If at all possible, could you please point me in the right direction, or at least tell me which pipeline would be best suited for my needs?

ExampleVitra2.cppipe (15.3 KB)


This is not a trivial assay, so don’t beat yourself up about not coming up with a solution :smiley:

I put together a pipeline which should get you started. The key approach I used here was to use the “Dark holes” feature in the EnhanceOrSupressFeatures module; this has the effect of highlighting dark features surrounded by bright rings. So I used this once to extract the sheath cells themselves (plus all of the other cells) and then again to get the vein. Then I identify the sheath cells as those objects touching the enclosing vein.

assay.cppipe (15.3 KB)

Hi Mark,

Thank you so much for your help. :smile: I was able to run the pipeline. As to the thorniness of the challenge that you hinted to, the results produced a good degree of false positives and negatives. Once the software outputs the final image, is there a way for me to go back and manually unselect the false positives and then select the false negatives? I was doing some investigating on the forums, and I came across one post where david suggested using ilastik? Do you think that software would be appropriate in my case?