Counting muscle bundles

Hi Cell Profiler family

My name is Chinedu Nworu and I am a graduate student at the University of Arizona. I am having problems with trying to get cell profiler to identify muscle bundles (objects) in a given set of images.

I currently have a decent pipeline (Sam_truncated02.cp) but it is not able to consistently detect the objects I am looking for. It seems to be an intensity value threshold issue that I cannot figure out.

  1. The pipeline takes color images, converts them to grayscale, illumination corrected, image is then inverted, the smoothing module was used to remove cracks (explained in detail in point 2) and then muscle bundles would be identified.
  2. The pipeline was initially able to detect individual bundles but would occasionally group 2-3 bundles together and call it one. We also had an issue with the cracks. These cracks are found within each bundle. If deep enough and close to the edge of a bundle, the pipeline would misconstrue that for an edge and then split a single muscle bundles into two.
  3. My question is, can you fine tune my current pipeline or give me some clues on how to increase the intensity differences between muscle bundles so when I identify objects, it doesn’t lump several individual objects as one? Or, is there a way we can increase the pixel intensity values around what we consider edges of an object?

Attached are some images and the pipeline.
Thanks a lot, Chinedu





Sam_truncated02.cp (7.32 KB)

Hi Chinedu,

I’m attaching a modified pipeline which should point you in the right direction; look to the annotations at the top of the module settings for more on my thought process. The main recommendation I have to continue adjusting the smoothing and maxima suppression settings in the 2nd IdentifyPrimaryObjects module in order to get the proper segmentation.

Regards,
-Mark
2011_08_20.cp (7.01 KB)