I’m trying to construct a pipeline which very accurately outlines irregular cell shapes taken in brightfield images. The ultimate intention of this pipeline is to classify macrophage polarizations (M0, M1, M2), so capturing shape oddities such as small, thin projections off the edges of the cell and long extended ‘arms’ is important. My attempt at this pipeline generally captures the crude shape of and segmentation between the cells, but does not have nearly the precision I’m aiming for, especially at the edges. It also tends to over-segment the cells despite my having manually set de-clumping values. For instance, is there some way to capture the level of outline detail that’s shown in the three images below, in which the cell outline lies within the halo and hugs the cell body? I’d love to hear any suggestions about effective modules I might use towards these ends.
Below is my pipeline along with three representative images I’m working with.
Thanks so much for any suggestions,
Test.cpproj (650.2 KB)