Adapting custom classifications from outside CP into CP classification rules

cellprofiler
classification
cellprofiler-analyst

#1

Hi,

I’m looking for advice for how to adapt custom classifications of objects into rules that can be used by Cell Profiler.

We used CellProfiler Analyst to classify different spheroid morphologies from our images. We take these rules and use them directly in CP to classify objects based on these rules made in CellProfiler Analyst. This worked very well to do user-defined classifications in CP as defined by CPA rules. We then import these classification rules into CP to classify cells and overlay outlines (colour matched to classification) onto our long-term time lapse movies.

Now, we want to do something additional. We take the primary measurements generated by CP and use external programs (e.g. KNIME and R) to generate data-driven detection of how many subtypes of cell behaviours there are from long-term time lapse movies. This works well.

What I would like to do is to have CP load these automated classification rules, and overlay classification outlines (similar to aforementioned_. Is there a way that I can workout how to take classification rules generated elsewhere and format them into syntax that CP can use to do this for us?

I understand the syntax of the the following example. What I can’t work out is how to give the weighting score.

IF (Nuclei_AreaShape_Area < 351.3, [0.79, -0.79], [-0.94, 0.94])

Any ideas?

Thanks very much!


#2

Hi David,

It looks to me like there’s a very complex and iterative process to determine the weights, so without knowing exactly what sort of classifier it is (Fast Gentle Boosting vs another type) and what the program you’re using spits out, it’s hard for me to make a more specific recommendation. Can you give a bit more intel?