Segmentation of retinal cells

My pipeline is not accurately segmenting the cells in my group’s images. I’ve used it before when analyzing images of mouse retina, and it worked pretty well, but my group uses a different staining method with our samples of human retinas that I’m trying to gather morphological data on. Manual tracing is being performed to assist with cell segmentation, but it’s too time consuming. Can anyone offer suggestions as to how I can modify my pipeline to reduce the extra green fluorescence that does not mark the edges of cells, or at least enhance the edges enough to make segmentation a little easier? I’ve attached one of my sample images, an image showing where the sample’s manually traced edges are, and my current pipeline.
pipe.cp (11.8 KB)




I’m attaching a pipeline which should get you closer to the goal (some more tweaking may be needed). The EnhanceOrSupressFeatures module can be used to remove the background fluorescence, leaving the cell edges, which assists in identifying the cells. However, the dividing lines may still be hard to discern in places. I’ve tried two approaches, both of which are included in the pipeline.

Cheers,
-Mark
2012_01_21.cp (13.2 KB)