Detection of focal adhesions - separation


I am trying to detect focal adhesions in multiple cells in a single field of view. I’ve got a pipeline that I’m pretty happy with; however, two major issues remain. I’ve tried reading old posts and changing many different parameters, but I can’t seem to get these just right.

  1. Identification of individual cells. Generally pretty good, using DAPI for identifyprimaryobjects and then an actin stain for identifysecondaryobjects. However, sometimes it identifies a small part of one cell as belonging to another. (See the second smaller cell from the top and the big cell in the topcenter as an example).

  2. The more major issue is the detection of focal adhesions. The program tends to group too many of the focal adhesions together. I’m attaching one example. It also sometimes doesn’t see the whole length of a focal adhesion, as they tend to get dimmer at the ends.

Can you suggest any changes to my pipeline that will result in a better separation of these focal adhesions? On the imaging end, I could try taking higher magnification images of single cells, but it would be nice (and much faster) to be able to quantify images containing multiple cells (which would have to be at lower magnifications).

I’ve attached an example set of images and my current pipeline.

A separate question I have is about saving the output images that pop up after each step of the pipeline. I know that those can be saved when they are open, but when I run multiple image sequences through a pipeline, it only keeps the current one open (briefly) or the final one open. It would be nice to be able to save all of the identifyprimaryobjects image windows so that I can check that things are running smoothly without running each image through individually. Please let me know if this is possible.

Thank you very much in advance for your response! I think this program will be extremely useful, and the help sections and this forum are very helpful in figuring out how everything works!

Hi Jillian,

Changing the method to identify the secondary objects (from the default “Propagation” to “Waterhed - Gradient”) seems to achieve better results, at least for this image set.

I see that you are using EnhanceOrSupressFeatures to increase the contrast prior to detection. I think this the right approach, but with top-hat filtering (which is what this module is doing with “Speckles” as the feature), the filter size should be roughly the size of the feature you want to preserve. Since you are looking for long, thin features, I would reduce the feature size to the approximate width of the FAs (maybe about 3-5 pixels?). Doing this, plus reducing the smoothing filter size in IdentifyPrimaryObjects (maybe 2 pixels) in order too much blurring (and hence merging) of these finer features, might get you closer to the results you want.

Our recommendation is collect the outlines of each set of objects (available from the “Retain outlines…” setting), use OverlayOutlines to superimpose them on an image of choice, and then save the overlay image using SaveImages.

Hope this helps!

Thank you very much!!

That worked great. I think that lowering the value for the EnhanceFeature feature size to 3 made the FAs much clearer. I also decreased the smoothing filter as you suggested.

I will try saving the image outlines to test that it all works smoothly for a large batch of images.

Thanks again!

  • Jillian