Can you upload some example images, with an explanation of what exactly you want to extract from them?

Dear Lambert,

We are trying to develop new methods for cell-cell contact identification for live cell images and I need some existing methods to study. Which might be helpful to design new methods. Could you please provide any literature in this regard.

I do image-processing and don't know much about synapses, tight junctions, phagocytosis or whatever cell-cell contact you are studying.

If the information can easily be extracted from the images, there is hardly any need to read a lot of literature, nor is there then a need to improve existing methods...

So, can you upload some example images, with an explanation of what exactly you want to extract from them?

Dear Lambert,

Thank you very much for the paper you have provided. We are trying to find out the cell-cell communication/contact in bladder cancer mono layer cell lines. We also trying to figure-out the cell boundaries in the live cell images. Hence, develop new methods to quantify cell boundaries which might help to understand the cell-cell contact.

Dear Chandrabhan,

Interesting, but I'm still waiting for images (~!@#$%) ...

The objective is 'understand the cell-cell contact'; excellent!

A couple of image-processing steps is just a means; a method. I fail to see why you want to develop new methods when existing methods might be adequate.

(I don't know -haven't seen images yet- but) IF segmentation is easy, I wouldn't even bother searching the literature to know what other research groups did.

Dear Lambert,

I am not much aware about the existing methods of cell-cell contact and segmentation analysis from the live cell images. Could you please provide some details, it might be helpful for me to understand existing methods.

Dear Lambert,

I have just started working in this field.

Dear Chandrabhan,

Please let me see some images!

Only then, and with an explanation of what you want to extract from them, I MIGHT be able to talk sense... Is it electron microscopy or what? I still have no clue.

So, press the 'Add attachment' button, select a high-resolution image, and we will go from there...

Hi,

This is an image from Phase Contrast Microscopy. I am not able to add more images. We are looking cell-cell contacts.

One method relies on watershed.

1. Denoise your image: try gaussian blur 1-2 pixel radius

2. Threshold Mask: try Otsu's method

3. Watershed: run watershed operation on copy of the mask

4. Difference: Subtract watershed mask from original mask

5. Highlighted lines should be indicative of identifiable cell-cell contact zones.

6. Use analyze particles to get measurements from these contact zones

Warnings:

Requires good starting image (avoid signal saturation)

Verify thresholding accuracy

Identify whether watershed separates all touching cells

Tweak as necessary

Notice in the attached image that red lines show watershed contact zones. Two contact zones, indicated by yellow arrows, failed to watershed correctly.

Possibly Lambert has another method, or can optimize the watershed separation

Seems fine to me Kristopher!

Probably watershed failed because of saturation of the Red channel. I inspected the two that failed; there was very little (but just enough) signal left, but I did not do the de-noising step.

To the north-east of the left-most yellow arrow, a cell-cell contact was missed. Here watershed could not do it because of saturation.

In such cases you could look for concavities: convert to nicely interpolated line segments at the threshold value (marching cubes in 2D) and calculate angular deficits.

Better is to reduce exposure time.

Cell-cell contacts are extremely difficult to assess with phase contrast microscopy. At the very minimum you should use a membrane marker like the membrane localized probes from clonetech. The most robust way to measure tight cell-cell contacts is to use the GRASP technology (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3424517/) which is based on fluorescence complementation with two cells expressing different halves of the split GFP. Note that cell-cell contact is very different from cell-cell communication. If you are interested in gap junctional communication, you should use a gap-junction permeant dye like lucifer yellow (http://www.ncbi.nlm.nih.gov/pubmed/11927380). Typically the strategy involves mixing cells labeled in a non-transferable blue dye with cells cotaining lucifer yellow. If the blue cells appear yellow after contact you know that cytoplasmic communication has taken place.

For actual cell-cell contact phase is indeed not optimal. Either dyes should be used as suggested from the previous post or you can try darkfield. The darkfield imaging modality will be a lot better suited than phase contrast to find edges and possibly connections between cells while phase contrast is probably better for coarser cell segmentation.

Hi, Kristopher,

Could you please let me know which technique or tool you have used to determine the cell-cell contact in the provided image? IT seems good !! Actually, I am looking cell-cell contact in unlabeled live cell images.

Chandrabhan

I engineered the solution above off the cuff. There is no formal description that I am aware of. Thus if you are looking for a name to the method or reference there isn't one. Otherwise simply follow the steps outlined and shoot me a question if you hit problems

Hi,

Best for cell-cell contact as an experiment is a dye transfer assay. If you are looking at closer contacts.

A microinjection with dye into cells and transfer to the next cell can be visualized beautifully and imaged too. dyes like neurobiotin. we use such approach in lab. But yeah post image processing or quantification, I am not sure. We take many images in confocal microscopy if there is junction protein you want to study.

Thanks for your answer but unfortunately I am looking label free methods to address cell cell contacts in live cell microscopy.