Active cells/contours and measures

[Magali SOYER]

Hello everybody,
I'm trying to measure fluorescence intensities of proteins in individual cells, on images comprising several cells that can be as groups.
I've tried to delineate the cells using the active cells and active contours plug-ins and then ROI measures; but I'm quite unsuccessful in separating the different cells.
Here is an example of a 4 channels pictures. I'm interested in the quantification in channel 3.

Can anybody help me with the segmentation?


Thanks a lot,


Magali

[Alexandre Dufour]

Hi Magali,

I've just had a look at your images, and it is clear that in their current state, your images are *quite hard* to analyze. In a nutshell, here are a few comments:

 1) the membrane labeling that you are using is not homogeneous, causing the membrane to appear bright in some locations and dimmer in other places, such that the contours cannot correctly be attracted to the "correct" boundaries. Ad-hoc methods could force an arbitrary separation, but they would be highly inaccurate. Which staining are you using? This looks like phalloidin, yet the multiple dot in the cytoplasm are a bit strange... Do you think you can improve the staining part at some point? Or changing it with another marker altogether?

 2) cells are closely confluent, causing them to overlap in many occasions. This is an extra issue that is hard to solve (at least for now), but I consider the issue as minor compared to the first.

I'm sending you here a protocol of what I've come up with (You can open it using the "Protocols" plugin). It starts by segmenting the nuclei (using channel 0), and grows an active contour starting from each nuclear membrane until it reaches bright intensity (considered as membrane, using channel 2, channel 3 gives worse results). As you run the protocol I think you will better understand what I mean by *quite hard*...

[Magali SOYER]

Hi Alexandre,
Thank you a lot for your answer. Indeed it's quite hard to analyze, even with the protocol you gave me.
To answer your questions:
1/ yeah the actin labeling (channel 3) is a bit strange, maybe I could find better pics. In channel 2, we are labeling a membrane receptor, so I'm not surprised it's not homogeneous. I will see if we can have another proper membrane staining, maybe with lectins.
2/ those cells are colon cancer cells and they tend to grow as groups, so even if not fully confluent, they might always form "junctions".

I will try to run the protocol on other images and come back to you to give you a feed-back.

Thanks again,

Magali