Manual counting of cells in a z-stack with maximum intensity projection - question regarding pixel intensity

I’ve outlined my underlying data set in a separate post. I am now looking to take the same data and create a “gold standard” for comparison. The idea is to make a MIP image of the red (dead cells) and green (live cells) channels and then count them manually, with an additional channel merge image to clarify which cells appear alive on the green channel but are actually dead/dying by looking for colocation of the red channel.

Regarding pixel intensity. I have problems with my z stacks on higher cellular density images. That is the confocal wasn’t able to penetrate as well deeper into the stack. However, when I “window” the image well enough I can in fact see everything. So, my question is does this matter? If my understanding of MIP is correct, it doesn’t matter how terrible the penetration is - if the cell is even just barely visible it should show up well on the MIP. Am I missing something?

It will matter if you are attempting to apply a threshold, and whether or not it matters in the MIP depends on your MIP and other processign steps. If your noise is higher at the bottom of the stack than your signal is at the top of the stack, that can be a huge problem in general, no matter what you want to do.

Your linked thread shows very few cells, so if that is true in all of your images, you might be fine for a binary determination (red vs no red). If you want to say “this much red is too little, but this much red is enough,” then that will be a problem. The moment you start any kind of thresholding, you can run into problems, because weakly positive cells that are “high” in the Z stack could start to disappear, even with the MIP.

Whether all of the conditions are met for your measurements to be accurate, I can’t really say. If a whole cell is just barely visible in the stack (at the top), it will still be barely visible in the MIP, that does not really change much. All the MIP does is keep the brightest pixel in the Z stack at every location for each channel.

That meandered a bit as I kept getting distracted away from the computer, hopefully it ended up clear enough?