Focal adhesion segmentation

I am trying to analyse the size and intensity of focal adhesions (FAs) in my samples and am having a hard time segmenting individual FAs. Plus my pipeline misses a few of them. In addition my pipeline also picks up actin stress fibres at the edge of the cell elongating the FAs. I would like to avoid this from happening but may be impossible. I was wondering if you could help. I am also posting examples of starting and analysed images to stress improper segmentation and the pipeline used.
Just to give an idea of things I have already tried:
In the identify primary objects module:
I have tried smoothening filter size from 1-20 instead of auto…did not help
I have also tried the same filter size for " minimum allowed distance between local maxima"…did not help
In the enhance supress module:
I have tried using enhance neurites instead of speckles…helped a little bit in segmentation but ended up identifying too many actin tails
Also tried using enhance neurites and enhance speckles one after another and vice versa …did not help.

I do not not know how to manually add segments …would really appreciate your feedback!!


FAidentificationfinal.cp (9.14 KB)




Hi,

(Sorry for the delay in responding)

I don’t think I have too much to add to your pipeline; it’s a tough pipeline your posing. I’m attaching a pipeline that might do a little better. I’ve manually set the smoothing filter size and maxima distance threshold, as well as the thresholding method. The only other thing I can think of is to use MeasureObjectSizeShape to measure the object morphology and then use FilterObjects to remove objects that fall outside of a morphological criterion on your choice to distinguish actual FAs from stress fibers. For example, if the fibers are elongated, you might use the form factor as a suitable measure, where elongated objects have a form factor closer to 0 than those that are circular. Other measures might serve just as well.

If it turns out that you do need to add objects to the image, you can insert the IdentifyObjectsManually module into your pipeline. But of course, an solution that requires the least amount of manual intervention is always best. :smile:

Cheers,
-Mark
2013_04_02.cp (4.21 KB)