Feature recommendations for Weka Segmentation?

Above is a pre- and post-processing image of the GFP channel for my A549 cells.

Moi 5 HA rep 1 c8C1.tif (1.2 MB)

Above is a zoomed example image (orriginal) of my Alexa Fluor 647 channel corresponding to my RNA FISH probe puncta.

Background:

I’m fairly new to microscopy and image processing and especially new to Weka. I am limited to what I can experiment on my own, since I have limited access to a scientific computer that can handle the memory requirement of Weka. So, any and all advice would be much appreciated!

I’m working with the A549 cell line and performing RNA FISH experiments. I would like to answer two questions:

  1. What percent of cells are GFP+?
  2. Of the GFP+ cells, how many puncta are there?

Problem:

  1. In segmenting my cells, I would like to discriminate objects by size to avoid considering artifacts and debris as cells, but I have been unable to conclude what features are appropriate for doing so. Any recommendations?

  2. If I were to introduce a counterstain in my experiment, perhaps within the FITC or BFP channels (both colors are utilized in the system already but not important for quantification) would I be able to utilize this for enhanced segmentation in Weka?

For biological image analysis, a nuclear marker is usually necessary as a seed point for cells, if you want to count cells. If you just have a lot of cytoplasm, it’s rather hard to tell where one cell ends and another begins. Nuclear markers don’t solve this entirely, but it does give you a starting point.

Cell membrane dyes are markers can also make telling where cell edges are a lot easier. The whole project sounds like something that would be better for specialized software like CellProfiler, though.