count numbers of cells in brightfield image that have 3 proteins

I have brightfield images of cells and floruoscent images as well (these fluoroscent channels shows the proteins the cells have in the.There are three channels-red, blue and green hence 3 proteins). . How can I count number of cells that have these proteins. The other issue I am facing is if you see the brghtfield image, the cells are very oddly shaped, have different levels of grey in them and some cross over the other, when I try to threshold them to binarize, some of them appear broken and I cannot use imfill directly( If I could use imfill, I would just intersect them and find overlapping area). Is there any solution to finding the number of cells that have the three proteins given these concerns?

Answer 1

[Image Analyst]

Make your RGB image from the 3 channels

rgbImage = cat(3, redChannel, greenChannel, blueChannel);

then bring up the Color Thresholder on the Apps tab of the tool ribbon and threshold for purple/pink regions. Then tell it to Export the code.

To fill the blobs, try

mask = bwconvhull(mask, 'objects');

[AAS]

Thank you so much. I tried the conv hull for the brightfield image. This is what I get- Is there anything I can do about the circled regions?

IMAGE A: CONV HULL OF BRIGHT FIELD IMAGE:

IMAGE B: SIMPLE IMBINARIZE OF BRIGHT FIELD IMAGE:

IMAGE C: COLOR THRESHOLDED IMAGE FROM RGB IMAGE

IMAGE D: RGB IMAGE BEFORE THRESHOLDING

Basically If I can get Image A to work, I can probably do a Image A +Image C and count number of objects. Do you have any other suggestions?

[Image Analyst]

Why does it matter? You're never going to be 100% accurate. Why don't you just measure the area fraction and be done with it? You're never going to get every single pixel correct with such a grainy image. I think the area fraction would correlate well with whatever you're doing, like how many bacteria get killed with your antiseptic or whatever.