Detection of "elliptical rings" from microscopy images

Hi everyone,

I have a question regarding the segmentation of "elliptical rings" from microscopy images.

I attached an example here (as png) - these are NIH3T3 cells, which express a marker for the nuclear lamina - as visible by an "elliptical ring" (you can see 5 rings easily by eye, a sixth one appears upon increasing brightness/contrast). The other features in the image (i.e. the bright spots, or the signal around and within the rings) are not of interest at all.

To analyze the data, I need to measure various properties of this ring, but I'm already failing in the proper detection of those structures. I approached it with various thresholding and edge detection methods, but no combination resulted in anything useful for me. Since I'm only an amateur with image analysis, maybe someone could give me advice here.

As a result, I'd like a processed image which only shows the rings with a one-pixel border. Further, I will need to track these rings over time, but let's start with identifying them first.

Thanks in advance, I'm grateful for any help.

Best wishes, Matthias

Answer 1

[Swarooph]

Have you had a chance to take a look at this example?

Detecting a Cell Using Image Segmentation

It talks about various fundamental image processing techniques (edge detection, morphological operation etc.) to segment a cell. I would start here and try to understand how the code works for an example image. And then replace the example image with your image and play with the parameters to get it to appropriate values to see what you can make of it. For e.g. I blindly stuck your image into the code and changed the sobel fudge factor to 1.75 to get a decent edge detection. After this, you can play with other morphological operations to remove disturbances and sharpen the edges. The function bwmorph with the skeletonize operation (documentation here) might help in thinning out the lines.